Question 1 of 20

What does a digital camera ultimately measure before JPEG compression?

Question 2 of 20

Increasing ISO without changing shutter or aperture primarily:

Question 3 of 20

In discrete 2D convolution, flipping the kernel before sliding is required because:

Question 4 of 20

A separable Gaussian filter (1D horizontal then 1D vertical) is preferred because:

Question 5 of 20

The Canny edge detector includes non-maximum suppression to:

Question 6 of 20

Converting RGB to Lab color space before measuring color difference is useful because:

Question 7 of 20

Histogram equalization on a single grayscale image:

Question 8 of 20

Bayer demosaicing interpolates missing color channels because:

Question 9 of 20

Zero-padding an image before convolution mainly affects:

Question 10 of 20

Sobel operators approximate image gradients by:

Question 11 of 20

Shot noise in bright regions scales approximately as:

Question 12 of 20

Normalizing images to mean 0 and std 1 before CNN training helps because:

Question 13 of 20

The gradient magnitude √(Gx² + Gy²) is often used instead of Gx alone because:

Question 14 of 20

Gamma encoding in sRGB JPEGs exists primarily to:

Question 15 of 20

Hysteresis thresholding in Canny uses two thresholds to:

Question 16 of 20

White balance correction adjusts images because:

Question 17 of 20

Applying a large Gaussian blur before Sobel gradients helps because:

Question 18 of 20

Full well capacity limits:

Question 19 of 20

HSV separates hue from saturation and value, which is useful for:

Question 20 of 20

Double thresholding without hysteresis in Canny would likely: