Image Matting Quiz

Answers and explanations appear as you take the quiz.

1. What does the alpha matte represent in image matting?
   • The brightness of each pixel
   • The depth of each pixel
   • The object's category label
   • The opacity (foreground fraction) of each pixel between 0 and 1
2. How does matting fundamentally differ from binary segmentation?
   • Matting only works on video
   • Matting estimates fractional opacity for soft edges, not just hard masks
   • Matting produces hard 0/1 masks only
   • Matting ignores color information
3. What is a trimap in classic matting?
   • A three-color photo filter
   • A depth sensor output
   • A map marking definite foreground, definite background, and unknown regions
   • A type of camera lens
4. The compositing equation I = alpha*F + (1 - alpha)*B is underdetermined because:
   • There are more unknowns than the three known RGB values per pixel
   • It has no solution ever
   • Alpha is always exactly 0.5
   • Foreground and background are identical
5. What is notable about models like MODNet and Robust Video Matting?
   • They require a hand-drawn trimap for every frame
   • They perform trimap-free matting, often in real time
   • They only work on black-and-white images
   • They output 3D meshes
6. Which loss is commonly combined with alpha prediction error in deep matting?
   • A speech-recognition loss
   • A reinforcement reward
   • A classification cross-entropy on labels
   • A compositional loss that re-blends the prediction and compares it to the original image
7. Why might a deep matting network include a refinement stage after an initial alpha prediction?
   • To improve fine boundaries such as hair that a coarse prediction may miss
   • To replace every fractional alpha with a hard label
   • To discard the foreground color before compositing
   • To force all background pixels to become foreground
8. Which video-conferencing operation directly uses a predicted alpha matte?
   • Estimating the participant's spoken words from the waveform
   • Selecting an audio codec for the microphone
   • Blending a participant with a replacement background along soft edges
   • Assigning the call's network packets to a router