I85PREP
⌘ K
CG SET study materials
CG SET study resources

CG SET practice questions

One question at a time. Build your CG SET understanding.

First 50 of 200 questions. Each one lists the choices, the correct answer, and an explanation. The checker at the top of this page saves progress in this browser when it is available.

Continue in the practice tool above to load more questions, 50 at a time.

Question 1Graphics Mathematics and Geometry

A 3D vector v has components (3, -4, 12). What is its Euclidean length?

  1. 17
  2. 12
  3. 13 Correct answer.
  4. 19

Explanation

The length is the square root of the sum of squared components: 9 + 16 + 144 = 169, and the square root of 169 is 13. This is the standard Euclidean norm used throughout graphics transforms and normalization.

Question 2Graphics Mathematics and Geometry

In a right-handed coordinate system, the cross product a × b produces a vector that is:

  1. perpendicular to both a and b Correct answer.
  2. coplanar with a and b in every case
  3. the sum of a and b
  4. parallel to a but not to b

Explanation

The cross product is defined so its result is orthogonal to both input vectors, with direction set by the right-hand rule. Its magnitude equals the area of the parallelogram spanned by a and b.

Question 3Graphics Mathematics and Geometry

A point (x, y) is rotated counterclockwise by angle θ about the origin. Which matrix, applied as M·(x, y)ᵀ, performs this rotation?

  1. [[cosθ, sinθ], [sinθ, -cosθ]]
  2. [[cosθ, -sinθ], [sinθ, cosθ]] Correct answer.
  3. [[cosθ, sinθ], [-sinθ, cosθ]]
  4. [[sinθ, cosθ], [-cosθ, sinθ]]

Explanation

Counterclockwise rotation uses the matrix with cosθ on the diagonal and -sinθ in the upper-right, sinθ in the lower-left. Applying it to (1,0) yields (cosθ, sinθ), confirming the direction of rotation.

Question 4Graphics Mathematics and Geometry

What does a 4×4 homogeneous transformation matrix with a nonzero translation in its last column and an identity 3×3 upper-left block represent?

  1. A perspective projection
  2. A pure rotation
  3. A uniform scale about the origin
  4. A pure translation Correct answer.

Explanation

In homogeneous form the upper-left 3×3 block holds linear terms and the last column holds translation. With an identity block present, only the translation applies to points.

Question 5Graphics Mathematics and Geometry

The dot product of two nonzero vectors equals zero. What does this imply geometrically?

  1. The vectors have equal length
  2. The vectors are parallel
  3. The vectors are orthogonal Correct answer.
  4. The vectors are antiparallel

Explanation

The dot product equals |a||b|cosθ, so a zero result with nonzero magnitudes forces cosθ to be zero. That means the angle between the vectors is 90 degrees.

Question 6Graphics Mathematics and Geometry

Which matrix represents a uniform scale by factor s applied to 2D points in homogeneous coordinates?

  1. [[1,0,0],[0,1,0],[0,0,s]]
  2. [[s,0,0],[0,s,0],[0,0,s]]
  3. [[s,0,0],[0,1,0],[0,0,1]]
  4. [[s,0,0],[0,s,0],[0,0,1]] Correct answer.

Explanation

Uniform scaling multiplies x and y equally by s while leaving the homogeneous w coordinate at 1. The bottom-right entry must remain 1 to preserve point representation.

Question 7Graphics Mathematics and Geometry

In the standard pinhole camera model, what does the focal length control?

  1. The field of view and projected image scale Correct answer.
  2. The camera's world position
  3. The color depth of the framebuffer
  4. The near clipping distance

Explanation

Focal length relates the image plane distance to the projection, so a larger focal length narrows the field of view and magnifies the image. It does not set camera position or clipping planes.

Question 8Graphics Mathematics and Geometry

Two lines in a 2D plane are represented by homogeneous line vectors l1 and l2. How is the intersection point computed?

  1. The dot product l1 · l2
  2. The matrix product of their transposes
  3. The sum l1 + l2
  4. The cross product l1 × l2 Correct answer.

Explanation

In projective geometry, the intersection of two lines is the cross product of their homogeneous coefficients. This is the point-line dual of computing a line through two points.

Question 9Graphics Mathematics and Geometry

A triangle has vertices A(0,0), B(4,0), and C(0,3). What is its area?

  1. 5
  2. 6 Correct answer.
  3. 12
  4. 7

Explanation

The area of a triangle is half the base times height; here base 4 and height 3 give 4·3/2 = 6. Equivalently the cross product magnitude of AB and AC is 12, halved to 6.

Question 10Graphics Mathematics and Geometry

A 3D rotation matrix R is orthonormal. Which property must hold?

  1. R = Rᵀ with det(R) = 0
  2. RᵀR = I and det(R) = 1 Correct answer.
  3. RᵀR = I and det(R) = -1
  4. R² = I for every rotation

Explanation

Rotation matrices are orthogonal with unit determinant, so their transpose is their inverse and det(R) equals +1. The +1 determinant distinguishes rotations from reflections.

Question 11Graphics Mathematics and Geometry

Given the plane equation 2x - 3y + 6z = 12, what is the unit normal vector?

  1. (1/7, 1/7, 1/7)
  2. (2, -3, 6)
  3. (12, 12, 12)
  4. (2/7, -3/7, 6/7) Correct answer.

Explanation

The normal is the coefficient vector (2, -3, 6), whose length is 7. Dividing each component by 7 gives the unit normal.

Question 12Graphics Mathematics and Geometry

A quaternion q = (w, x, y, z) represents a rotation. What must be true for it to be a unit quaternion?

  1. w² - x² - y² - z² = 1
  2. w + x + y + z = 1
  3. x² + y² + z² = 1 with w = 0
  4. w² + x² + y² + z² = 1 Correct answer.

Explanation

Unit quaternions have magnitude one, meaning the sum of squared components equals one. This constraint keeps the represented rotation free of scale distortion.

Question 13Graphics Mathematics and Geometry

In barycentric coordinates (u, v, w) for a triangle, a point lies strictly inside the triangle when:

  1. u, v, w are all positive and sum to 0
  2. u, v, w are all negative and sum to 1
  3. u, v, w are all positive and sum to 1 Correct answer.
  4. u, v, w are all zero

Explanation

Barycentric weights of an interior point are positive and normalized to sum to one. Each weight expresses the relative contribution of a vertex.

Question 14Graphics Mathematics and Geometry

A point p is reflected across the plane with unit normal n passing through the origin. Which formula gives the reflected point?

  1. p - (p · n)n
  2. p + 2(p · n)n
  3. p - 2(p · n)n Correct answer.
  4. 2p - (p · n)n

Explanation

Reflection removes twice the component of p along the normal, giving p - 2(p·n)n. The tangential component is preserved while the normal component reverses.

Question 15Graphics Mathematics and Geometry

Which operation produces the inverse of a translation matrix T(t) in homogeneous coordinates?

  1. T(2t)
  2. T(t) with the last row zeroed
  3. T(-t) Correct answer.
  4. T(t)ᵀ

Explanation

Translating by t and then by -t returns to the origin, so the inverse is translation by the negated vector. The transpose of a translation is not its inverse.

Question 16Graphics Mathematics and Geometry

Two vectors a and b satisfy |a| = 3, |b| = 5, and the angle between them is 60 degrees. What is |a × b|?

  1. 15√3/2 Correct answer.
  2. 15
  3. 15/2
  4. 7.5√3/2

Explanation

The cross product magnitude equals |a||b|sinθ, giving 15·sin60° = 15·(√3/2). This value is the area of the parallelogram spanned by the vectors.

Question 17Graphics Mathematics and Geometry

A perspective projection matrix maps a point with view-space depth z to clip space. What happens to points at the near plane relative to points at the far plane after the perspective divide?

  1. Depth values become independent of z
  2. Near points map closer to -1 or 0 and far points compress toward the opposite bound Correct answer.
  3. Near points map to the far bound and far points to the near bound
  4. Near and far points map to identical depth values

Explanation

Perspective projection stores depth divided by w, producing a nonlinear mapping where near depths spread out and far depths compress toward the far limit. This nonlinearity causes z-fighting at large distances.

Question 18Graphics Mathematics and Geometry

Which statement correctly describes the relationship between a rotation matrix and its corresponding quaternion?

  1. Each rotation maps to exactly one quaternion
  2. Each rotation maps to four distinct quaternions
  3. Quaternions cannot represent rotations beyond 180 degrees
  4. Each rotation maps to two quaternions, q and -q Correct answer.

Explanation

Unit quaternions form a double cover of the rotation group, so q and -q encode the same orientation. This sign ambiguity is why interpolation must choose the shorter arc.

Question 19Graphics Mathematics and Geometry

What is the determinant of a 2D affine transform matrix whose linear part is a rotation by θ combined with uniform scale s?

  1. s
  2. s² Correct answer.
  3. cosθ
  4. 2s

Explanation

The linear part is s times a rotation matrix, and in 2D the determinant scales by s². Rotation contributes a determinant of one, so the result is s².

Question 20Graphics Mathematics and Geometry

In homogeneous coordinates, how is a point at infinity represented in 2D?

  1. A vector with w negative
  2. A vector with w = 0 Correct answer.
  3. A vector with w = 1
  4. A vector with all components equal

Explanation

Setting the homogeneous weight w to zero yields a direction rather than a finite position, representing a point at infinity. Finite points use nonzero w, typically normalized to one.

Question 21Graphics Mathematics and Geometry

A surface normal is transformed by a non-uniform scale matrix M. To keep normals perpendicular to the surface, which matrix should be applied to the normal?

  1. The transpose of M
  2. The inverse of M
  3. M itself
  4. The inverse transpose of M Correct answer.

Explanation

Normals transform by the inverse transpose so they remain orthogonal to transformed tangent vectors. Using M directly distorts normals under non-uniform scaling.

Question 22Graphics Mathematics and Geometry

A cubic Bezier curve has control points P0, P1, P2, P3. What is the tangent direction at the start point P0?

  1. Proportional to P1 - P0 Correct answer.
  2. Proportional to P2 - P0
  3. Proportional to P3 - P2
  4. Proportional to P3 - P0

Explanation

The derivative of a cubic Bezier at t = 0 is 3(P1 - P0), so the initial tangent points toward the first control handle. This property underlies the intuitive handle-based editing of curves.

Question 23Graphics Mathematics and Geometry

A ray with origin o and direction d intersects a plane defined by normal n and offset such that n · x = c. Which expression gives the ray parameter t at the intersection?

  1. (c + n · o) / (n · d)
  2. (c - n · o) / (n · d) Correct answer.
  3. (n · d) / (c - n · o)
  4. (n · o - c) / (n · d)

Explanation

Substituting x = o + td into n·x = c gives n·o + t(n·d) = c, which solves to t = (c - n·o)/(n·d). A zero denominator means the ray is parallel to the plane.

Question 24Graphics Mathematics and Geometry

A matrix M has eigenvalues 1, 1, and 0.5 with a full set of linearly independent eigenvectors. What geometric effect does repeated application of M produce?

  1. Convergence toward the plane spanned by the eigenvectors with eigenvalue 1 Correct answer.
  2. Rotation about the eigenvector with eigenvalue 0.5
  3. Unbounded growth in all directions
  4. Convergence toward the eigenvector with eigenvalue 0.5

Explanation

Components along eigenvectors with eigenvalue 1 are preserved while the component along the 0.5 eigenvector decays. Repeated multiplication therefore projects points toward the invariant plane of the unit-eigenvalue directions.

Question 25Graphics Mathematics and Geometry

Why can gimbal lock occur with Euler angle rotations?

  1. The rotation matrix becomes non-orthogonal
  2. Two rotation axes become aligned, removing one degree of freedom Correct answer.
  3. Quaternion normalization fails
  4. The determinant of the rotation matrix drops to zero

Explanation

At certain angle combinations two of the three rotation axes coincide, so one rotational degree of freedom is lost. This singularity motivates quaternions or rotation matrices for robust interpolation.

Question 26Graphics Mathematics and Geometry

In homogeneous coordinates, a 2D point (x, y) is represented as (x, y, 1). What does the third coordinate w typically represent when w ≠ 1?

  1. The z-coordinate of the point in 3D space
  2. A scaling factor that must be divided out to recover Cartesian coordinates Correct answer.
  3. The distance from the origin to the point
  4. A flag indicating whether the point is at infinity

Explanation

In homogeneous coordinates, a point (x, y, w) with w ≠ 0 corresponds to the Cartesian point (x/w, y/w). The w coordinate acts as a scaling factor; dividing by w recovers the standard 2D coordinates. When w = 0, the point is at infinity.

Question 27Graphics Mathematics and Geometry

A 2D rotation matrix R(θ) is applied to a vector v. Which property must R(θ) satisfy to preserve lengths and angles?

  1. R(θ) is diagonal
  2. R(θ) is orthogonal with determinant 1 Correct answer.
  3. R(θ) has determinant -1
  4. R(θ) is symmetric

Explanation

A rotation matrix in 2D is orthogonal (its transpose equals its inverse) and has determinant +1, ensuring it preserves orientation and distances. This makes it a special orthogonal matrix, which guarantees length and angle preservation.

Question 28Graphics Mathematics and Geometry

A triangle has vertices A(0,0), B(4,0), and C(0,3). Using barycentric coordinates, what are the coordinates of the centroid?

  1. (0, 1/2, 1/2)
  2. (1/4, 1/4, 1/2)
  3. (1/2, 1/2, 0)
  4. (1/3, 1/3, 1/3) Correct answer.

Explanation

The centroid of a triangle in barycentric coordinates is always (1/3, 1/3, 1/3), as it is the average of the three vertices. This holds regardless of the triangle's shape or position.

Question 29Graphics Mathematics and Geometry

A 3D point p = (1, 2, 3) is transformed by a translation matrix T that moves points by (4, -1, 2). Using homogeneous coordinates, what is the resulting point?

  1. (4, -1, 2)
  2. (5, 1, 5, 0)
  3. (5, 1, 5)
  4. (5, 1, 5, 1) Correct answer.

Explanation

In homogeneous coordinates, a 3D point is represented as (x, y, z, 1). Applying the translation matrix adds the translation vector to the point's coordinates, yielding (1+4, 2-1, 3+2, 1) = (5, 1, 5, 1). The w component remains 1 for points.

Question 30Graphics Mathematics and Geometry

A 3D rotation is represented by a unit quaternion q = (w, x, y, z). Which formula correctly computes the rotated vector v' of a vector v?

  1. v' = q + v
  2. v' = q v q
  3. v' = q v q* Correct answer.
  4. v' = q* v q

Explanation

For a unit quaternion q, the rotation of a vector v (embedded as a pure quaternion (0, v)) is given by v' = q v q*, where q* is the conjugate of q. This sandwich product correctly applies the rotation and preserves vector length.

Question 31Rasterization and Rendering Pipeline

In the rasterization pipeline, which space does the viewport transformation map normalized device coordinates into?

  1. Object coordinates
  2. Window (screen pixel) coordinates Correct answer.
  3. Eye (camera) coordinates
  4. Clip coordinates

Explanation

The viewport transform scales and translates NDC values (after perspective divide) into window coordinates measured in pixels and depth range. It is the final coordinate mapping before primitive assembly and rasterization. Object, eye, and clip spaces all occur earlier in the vertex pipeline.

Question 32Rasterization and Rendering Pipeline

What is the primary purpose of the depth (Z) buffer during rasterization?

  1. Resolving visibility by keeping the nearest fragment per pixel Correct answer.
  2. Holding the stencil mask for clipping
  3. Storing the final shaded color per pixel
  4. Accelerating texture filtering operations

Explanation

The depth buffer stores a per-pixel depth value so that a fragment is written only if it is closer than the stored depth under the default less-than test. This resolves which surface is visible without sorting primitives. Color storage and stencil masking are separate buffers.

Question 33Rasterization and Rendering Pipeline

A triangle's vertices are given in clockwise order on screen, and the pipeline is set to cull back faces with counter-clockwise front faces. What happens to this triangle?

  1. It is clipped against the near plane only
  2. It is rendered as a front face
  3. It is rendered twice, once per winding
  4. It is discarded as a back face Correct answer.

Explanation

Face culling classifies triangles by signed screen-space area, which reflects vertex winding. With CCW defined as front-facing, a clockwise triangle is back-facing and is discarded before rasterization. Clipping and double rendering are unrelated to winding classification.

Question 34Rasterization and Rendering Pipeline

Which statement correctly describes the rasterization of a triangle's interior fragments?

  1. Fragments are generated for every pixel in the screen
  2. Fragments are generated only at the three vertices
  3. Fragments are generated for pixel centers covered by the triangle Correct answer.
  4. Fragments are generated along the triangle's edges only

Explanation

Triangle rasterization determines which pixel sample points (typically centers) fall inside the triangle and generates a fragment for each. Interior coverage, not the whole screen or just edges, defines the fragment set. Vertex-only work belongs to the vertex stage.

Question 35Rasterization and Rendering Pipeline

Which transformation sequence correctly converts a model-space vertex to clip space in a standard graphics pipeline?

  1. View matrix, then model matrix, then projection matrix
  2. Model matrix, then view matrix, then projection matrix Correct answer.
  3. Model matrix, then projection matrix, then view matrix
  4. Projection matrix, then view matrix, then model matrix

Explanation

Vertices are transformed from model to world space by the model matrix, then to eye space by the view matrix, then to clip space by the projection matrix. Matrix order matters because these transforms do not commute. Applying projection before the view transform would place the camera transform in the wrong space.

Question 36Rasterization and Rendering Pipeline

What does the perspective divide accomplish after the projection transform?

  1. It divides depth by the far plane distance
  2. It divides colors by alpha to produce premultiplied values
  3. It divides clip coordinates by w to produce normalized device coordinates Correct answer.
  4. It divides screen coordinates by the viewport width

Explanation

After projection, clip coordinates are divided by their homogeneous w component, yielding normalized device coordinates. This division produces the perspective foreshortening effect. It is unrelated to color premultiplication or viewport scaling.

Question 37Rasterization and Rendering Pipeline

Which buffer typically stores coverage masks used to implement multisample anti-aliasing?

  1. The vertex buffer
  2. The multisample color and depth buffers with per-sample coverage Correct answer.
  3. The stencil buffer alone
  4. A single-sample color buffer only

Explanation

MSAA uses multisample render targets where color, depth, and coverage are stored per sample, then resolved to a single color per pixel. This gives smoother edges than single-sample rendering. Stencil and vertex buffers do not hold per-sample coverage for antialiasing.

Question 38Rasterization and Rendering Pipeline

In Gouraud shading, at what stage is lighting computed?

  1. At every fragment using interpolated normals
  2. After the depth test on visible fragments
  3. Only at the triangle centroid
  4. At the vertices, then interpolated across the triangle Correct answer.

Explanation

Gouraud shading evaluates the lighting model at each vertex and interpolates the resulting colors across the primitive. Per-fragment lighting with interpolated normals describes Phong shading instead. This distinction affects highlight quality on coarse meshes.

Question 39Rasterization and Rendering Pipeline

A fragment passes the depth test with depth testing enabled and depth writes enabled. Which statement best describes the typical outcome?

  1. The fragment's depth replaces the stored value and its color may be blended into the color buffer Correct answer.
  2. The fragment is always discarded after the test
  3. The color buffer is cleared before the write
  4. The stored depth is left unchanged regardless of the test

Explanation

When the depth test passes and depth writes are enabled, the new depth value overwrites the stored one. The fragment's color then proceeds to blending and the color write. The color buffer is not cleared per fragment.

Question 40Rasterization and Rendering Pipeline

Why does enabling early-Z rejection improve performance in a fragment-heavy scene?

  1. Fragments failing the depth test are discarded before the fragment shader runs Correct answer.
  2. It eliminates the need for a projection matrix
  3. It compresses textures to save bandwidth
  4. It reduces the number of vertices submitted to the pipeline

Explanation

Early-Z performs the depth test before fragment shading so occluded fragments skip the expensive shader. This saves shading work when overdraw is high. It does not change vertex counts, texture formats, or the projection transform.

Question 41Rasterization and Rendering Pipeline

Which interpolation qualifier causes a varying to be interpolated with perspective correction across a triangle?

  1. The flat qualifier
  2. The noperspective qualifier
  3. The centroid qualifier
  4. The default smooth interpolation Correct answer.

Explanation

By default, varyings use smooth interpolation, which is perspective-correct because it interpolates in a perspective-aware manner using 1/w. The flat qualifier selects a provoking vertex value, and noperspective interpolates linearly in screen space. Centroid affects sample location, not perspective correction.

Question 42Rasterization and Rendering Pipeline

What is the effect of enabling alpha blending with source factor SRC_ALPHA and destination factor ONE_MINUS_SRC_ALPHA?

  1. The fragment color completely replaces the existing color
  2. The destination color is written into the alpha channel
  3. The fragment color is combined with the existing color weighted by the fragment's alpha Correct answer.
  4. The fragment is discarded if its alpha is below one

Explanation

This standard over-compositing blend computes src.rgb * src.a + dst.rgb * (1 - src.a), giving transparency proportional to alpha. Opaque replacement would require disabling blending. Alpha values do not automatically discard fragments.

Question 43Rasterization and Rendering Pipeline

In the standard OpenGL-style pipeline, which stage runs after primitive assembly and before the fragment shader?

  1. The compute shader, which replaces rasterization
  2. The vertex shader, which transforms vertices
  3. The geometry shader, which precedes primitive assembly
  4. Rasterization, which generates fragments from primitives Correct answer.

Explanation

After primitives are assembled and clipped, rasterization converts them into fragments that feed the fragment shader. The vertex shader runs earlier, before primitive assembly. Geometry shading, when present, occurs before rasterization rather than after it.

Question 44Rasterization and Rendering Pipeline

A scene shows severe stair-step aliasing on polygon edges despite a high display resolution. Which technique most directly addresses edge aliasing?

  1. Increasing the polygon count of the models
  2. Switching from Gouraud to flat shading
  3. Supersampling or multisampling with a resolve step Correct answer.
  4. Reducing the depth buffer precision

Explanation

Edge aliasing comes from binary coverage decisions per pixel, so sampling multiple subpixel locations and resolving them produces smoother edges. Adding geometry does not change per-pixel coverage. Shading model and depth precision do not remove geometric aliasing.

Question 45Rasterization and Rendering Pipeline

What is the purpose of the guard-band and clipping stage before rasterization?

  1. Computing vertex normals for lighting
  2. Removing primitives or portions outside the view frustum to avoid invalid rasterization Correct answer.
  3. Converting textures into mipmaps
  4. Sorting primitives front to back for the depth buffer

Explanation

Clipping trims geometry against the frustum planes so only valid portions reach rasterization, and guard-band clipping avoids unnecessary clipping for primitives near the screen. Sorting, mipmap generation, and normal computation are separate operations.

Question 46Rasterization and Rendering Pipeline

Which depth-test function renders a fragment only when its depth is less than the stored depth?

  1. LESS Correct answer.
  2. GREATER
  3. NEVER
  4. ALWAYS

Explanation

The LESS comparison passes when the incoming depth is smaller, meaning the fragment is nearer under a typical depth convention. GREATER passes for farther fragments, ALWAYS passes unconditionally, and NEVER always fails. The choice affects which surfaces remain visible.

Question 47Rasterization and Rendering Pipeline

Why can z-fighting appear when two coplanar surfaces are rendered with a distant far plane?

  1. The vertex shader receives duplicate vertices
  2. Limited depth buffer precision makes nearly equal depths indistinguishable Correct answer.
  3. Back-face culling removes both surfaces
  4. The color buffer overflows when both surfaces write

Explanation

Depth values are quantized to the buffer's precision, and a distant far plane compresses precision, so nearly equal depths can compare inconsistently and flicker. This is a precision issue, not a color or culling issue. Reducing the far/near ratio or using reversed-Z helps.

Question 48Rasterization and Rendering Pipeline

What does the viewport depth range mapping typically do to normalized device depth?

  1. It converts depth into a color value
  2. It leaves NDC depth unchanged
  3. It maps NDC depth from [0,1] to [-1,1]
  4. It maps NDC depth from [-1,1] to the buffer's [0,1] range Correct answer.

Explanation

The viewport transform scales and biases normalized device depth into the depth buffer's representable range, commonly [0,1]. This mapping is part of the same transform that maps x and y to window coordinates. It does not convert depth into color.

Question 49Rasterization and Rendering Pipeline

Which statement best explains why deferred shading can reduce lighting cost in scenes with many lights?

  1. It shades every fragment of every primitive before depth testing
  2. It computes lighting only at vertices
  3. It shades each visible surface once using stored G-buffer attributes Correct answer.
  4. It removes the need for a depth buffer

Explanation

Deferred shading first rasterizes geometry into a G-buffer of attributes, then applies lighting once per visible pixel, avoiding repeated shading of overdrawn fragments. It still needs depth information and does not move lighting to vertices. This makes many-light scenes cheaper than naive forward shading.

Question 50Rasterization and Rendering Pipeline

A texture lookup uses mipmapping with trilinear filtering. What does trilinear filtering interpolate between?

  1. The four nearest texels in one mip level only
  2. The depth values of neighboring fragments
  3. The vertex colors of adjacent triangles
  4. The two nearest mip levels, each bilinearly sampled Correct answer.

Explanation

Trilinear filtering samples bilinearly within each of the two adjacent mip levels and blends the results according to the level-of-detail fraction. This reduces shimmering compared with bilinear minification. It does not involve vertex colors or fragment depths.