Colour Processing, Pixel Operations Flashcards
(12 cards)
What is metamerism? (2024)
- Light is a combination of wavelengths.
- Different combinations may appear to be the same colour.
We apply the mapping u -> 0.7u + 12 on the pixel values of a grayscale image.
- How will this impact the image histogram and cumulative histogram?
- Does the mapping increase or decrease the perceived contrast of the image?
2024
- The operation is taking the original pixel, scaling down the intensity and then adding a brightness of 12.
- It will skew the more values to the left, assuming a histogram of intensity bins 0-255.
- No, as all the pixel intensity values are all being scaled and added the same. But depends on the image. If most pixelsare near 255, it might cause high contrast. More pixels are being saturated to 255 (white).
Briefly explain what metamerism is. Give an example.
2023
- Light is a combination of different wavelengths.
- Different combinations may be the same colour.
E.g. red + blue = purple
Briefly explain why SSIM is sometimes preferred to PSNR.
2023
SSIM = structural similarity index
- Introduced to predict the perceived quality of an image.
- Based on some of the HVS (i.e. Weber’s law and activity masking)
- Shown to perform better than standard metrics such as MSE or PSNR.
Why are, in the XYZ colour space, the different RGB gamuts always represented as triangles?
2023
- The RGB gamut is a triangle that fits inside the human gamut.
- Each combination of RGB values lie inside the triangle.
- The vertices of the triangle correspond to the 3 primary colours Red, Green and Blue.
Explain you could use colour information to extract a binary mask of the grapes in the picture below.
What are the possible pitfalls for a purely colour-based approach?
2023
- Plot the histogram for the three different colour channels - RGB.
- Blue would be the most useful colour channel to extract the grapes. This can be used to get the main features.
- Apply the threshold; mask_blue = uint8(pic > t_b). This applied a logical operator that makes values above t_b 0, preserving the colours of choice
In 2D signals and systems, when we talk about a LSI system, what does LSI stand for? What are the conditions for a system to be LSI?
2020
Linear-Shift Invariant
- Obeys linear superposition (i.e. f(ax+ by) = af(x) + bf(y)
- If we shift the input in time, then the output is shifted by the same amount.
How does the RGB colour space gamut compare to the human gamut?
2019
- RGB gamut is a triangle that fits inside the human gamut.
- Each combination of RGB values, lie inside that triangle.
- Vertices of the triangle correspond to the 3 primary colours (i.e, R,G,B)
What do the contrast sensitivity graphs tell us about the human sensibility to spatial frequencies? How are they exploited in the JPEG standard?
2019
What is the YUV colour space? (exact weights definition are not required). Why are using the YUV colour space in broadcasting.
2019
- YUV is used so that TV colour signals could be backward compatible with black and white TV sets.
- 1 luminance (Y), 2 chrominance values (U,V)
- The higher weight for green reflects our sensibility to the green wavelength.
To segment the green vegetation on a picture, your code applies a thresholding operation on the green channel of the RGN image (G > 150). Comment.
2019
150 seems to be a fine threshold, since green is the dominant colour in the colour space.
Watch out if the background has any green in it.
You receive a number of grayscale of CT scans. All images correspond to the same part of the body, but due to calibration issues, the gamma is not set consistently across images. You know that if the images were properly calibrated, you could isolate the feature of interest with a simple threshold 90 < Y < 120. What can you do to make your system work?
- Normalise the values.
2.
