Spatial domain vs. Frequency domain
Change the intensity of each pixel in order to enhance the image
Simplest form : Intensity transform with smallest box 1x1
Some basic gray-level transformation functions used for image enhancement.
Gamma Transformation : gamma > 1 : image getting darker
- Piecewise-Linear Transformation Function
S=T(r)
장점 : the form of function 임의적으로 complex할 수 있다.
Contrast stretching -> slope < 1, contrast down, slope > 1, contrast enhance
단점 : trade off 발생 - 가운데에만 contrast를 높일 수 있고 그것보다 작거나 크면 오히려 contrast가 줄어든다
: 즉, 모든 부분에서 contrast를 증가시킬 수 없다(not uniformly)
Full range linear stretching < - y=x in r1 to r2
Threshold (intensity transformation) function <- y=step(x), counting cells
Gray-level slicing : highlighting a specific range of gray levels,
- How can we preserve kidney and blood vessel?
input S=T(r) Angiogram
Slicing left intact : shape of the flow of the contrast medium to detect blockages
Gray leftA~B darker, other intact : actual flow of the contast mediumas a function of time
Bit -Plane slicing
mainly used for data compression and progressive transmission
MSB : b7 * 2^7
LSB : b0 * 2 ^0
b7, b6, b5 forms general shape(big storage information) and under that area gives details(less important)
plane 8~5만 sum해도 input이랑 굉장히 비슷해진다.
- 시험에 나오는 Histogram Processing
히스토그램은 무엇인가? : intensity는 각각 occurance를 몇개를 갖고 있냐에 대한 식
nomalized histogram : probability of occurences
히스토그램을 통해서 image가 밝은지 어두운지 contrast가 높은지 낮은지를 알 수 있다.
Desired image quality : broad and nearly uniformly distributed
'AI Engineering Topic > Digital Image Processing' 카테고리의 다른 글
Digital Image Processing ch4 Fourier Transform [1] (0) | 2022.04.12 |
---|---|
Digital Image Processing ch3 Smoothing & Sharpening spatial filter [3] (0) | 2022.04.12 |
Digital Image Processing ch 3 Histogram matching [3] (0) | 2022.04.12 |
Digital Image Processing Ch3 [2] Histogram Equalization (0) | 2022.04.06 |