point based filtering Flashcards
explain the point-based filtering
in this case, the output image depends solely on the value of the image pixel
what are the three types of filtering
point filtering where the computation relies on the values pixel of the image
region-based filtering where the values of the output image rely on its neighbourhood for instance kernel convolution and morphology
frequency base filter when the image is projected to the Fourier space, we filter on the Fourier space and then project it back to the Fourier space
what is an underexposed image
it is a an image with high contrast when we would notice a high number of the dark pixels and low number of light values
explain the effect of histogram equalization and how it can be computed
the idea consists of constructing a lookup table that stretches the picked gray levels meaning we are more interested to stretch higher range of gray level and compressing the low range of grey level
explain how the sensor acquire the colors
usually to acquire the colour the sensor uses 3 colours
what is the difference between Gaussian pyramid and subsampling pyramid
the applying a low pass filter after quantification while the other applies a direct quantification
what is level sampling
consist of limiting the number of image encoding, each time we encode the image with one bit less for instance
what is the entropy
entropy of an image reflects how many bits we need to encode a specific pixel gray level
what is the historgram of an image
reflects how many times we see a specific gray level or in another word the distribution of an image
what is the co-occurrence matrix
Describes the relative position of the gray level in the image
how does the window impacts the cooccurrence matrix
incresing the window mean the value in the pixel will be less related to its neighbourhood
what are the types of image compression
lossy compression and non-lossy compression
for images lossy compression is not an issue as the brain can easily cope with that
what is the idea behind huffman encoding
it consists of a way to encode common values with less
what is run length
usually only work binay images, instead of sending the image we send the value reflecting the
what is the idea behind image restoration
the idea is that we have the original image f that undergo a deformation H and additive noise N
