What is the study of digital image processing based on examples and codes?

Image processing (and machine vision) is a great research direction in school, and many graduate students and doctoral students are engaged in this research under the guidance of their mentors. In addition, in terms of work, there are indeed many jobs and opportunities to be found in this area. And this situation is becoming more and more prominent. So what's the problem with image processing, and today we're going to talk about it. The topic of image processing is very very wide, the extension is very far-reaching, the new topic is still emerging. The 12 major directions given below, which I think can be seen as a part of the basic field, and which are also interconnected

1 . Grayscale Adjustment of Image

Image gray histogram, linear transformation, non-linear transformation (including logarithmic transformation, power transformation, exponential transformation, etc.), gray-scale stretching, gray-scale equalization, histogram regulation, etc.).

For example, the histogram is regulated (see http://blog.csdn.net/baimafujinji/article/details/41146381for code)

CLAHE (Contrast limited adaptive histogram equalization) adaptive histogram equalization (from

http://www.cnblogs.com/Imageshop/archive/2013/04/07/3006334.html)

2, the geometric transformation of the image

Image panning, image mirroring, transpose, zooming, and rotation. This also includes interpolation algorithms (which are necessary for some geometric transformations), such as nearest neighbor interpolation, bilinear interpolation, and so on.

Geometric transformations are also closely related to the image's filter effects, and some of the effects are implemented in nature as geometric transformations of some kind. For example

3. Image effects and filters

There are many applications in this area, and you can think of the filters in Photoshop.

Document combining Sketch and Tone for Pencil Drawing production The effect of turning a natural image into a freehand sketch

such as Emboss effect

Stickers too annoying, more results please see http://blog.csdn.net/baimafujinji/article/details/50500757

4. Image enhancement

The content includes smoothing of images (simple averaging, median filtering, Gaussian smoothing, etc.) and sharpening (such as the Laplace method).

Many algorithms in the enhanced processing are coincident with the noise reduction algorithm in image restoration. Now the smoothing algorithm of preserving edge (or texture structure) belongs to the research hotspot. Skin-rejuvenation algorithms like those in a beauty camera are based on this. More common bilateral filtering (see http://blog.csdn.net/baimafujinji/article/details/41598455 for the code I gave)

TV denoising based on Total variational method (http://blog.csdn.net/baimafujinji/article/details/42110831), nonlinear diffusion denoising based on PM equation (http// blog.csdn.net/baimafujinji/article/details/42110831) and so on.

5. Image Restoration

Broadly speaking--image denoising, image de-fog, image blur all belong to this category

The denoising example is the effect of the median filter I implemented with Magichouse(http://blog.csdn.net/baimafujinji/article/details/50500757) to deal with salt and pepper noise. In addition, some noise reduction algorithms based on nonlocal mean are the hotspots of current research (e.g. BM3D, NLM, etc.)

Image blur (picture taken from my digital image processing principles and Practices (Matlab edition))

Go to the Fog code see (http://blog.csdn.net/baimafujinji/article/details/30060161) or reference my "Digital image processing principles and practices (MATLAB version)"

6 . Image compression and coding

Think about how BMP images are converted into jpg,jpg how do they become PNG? These are the contents of the image compression coding to be explored.

7. Edge detection and Contour tracking

Edge detection in image processing is an " old" topic, I do not give an example of specific. Here is an example of contour tracking

8. Image segmentation

You can assume that contour tracking is also a way to achieve image segmentation.

This is an example I gave in the principle and practice of Digital Image Processing (MATLAB edition)--Segmentation of potato images by watershed algorithm.

9. Morphological processing of images

This also belongs to a very old image processing way. including expansion, corrosion, refinement, hit/hit, open/close operation, etc. However, some applications that count granular objects are still very effective.

10, the frequency domain transformation of the image (or called orthogonal transformation)

Fourier, discrete cosine, Walsh-Hadamard transform, K-L (Calonan-Roy) transform (also called Hotelling Transform or PCA), wavelet transform (wavelet transform also has many kinds, such as Haar wavelet, Daubechies wavelet, etc.)

It doesn't make much sense to simply transform the frequency domain, it often has to be combined with specific applications to play a role. For example, image compression, embedding digital watermarking, image fusion, image denoising and so on.

For example, examples of image compression using PCA are shown in

http://blog.csdn.net/baimafujinji/article/details/50373143 (source code please see my blog)

In the example, using wavelet fusion to repair the image blur caused by focus failure (original left and middle images have some unclear, fusion becomes recognizable) source code visible

http://blog.csdn.net/baimafujinji/article/details/49642111

11. Image Fusion

In a broad sense, convergence contains at least three parts: like the above wavelet-based fusion We also know is a fusion, the other is to hide for the purpose of similar embedded fusion, the third is matting. Matting sometimes anti-righteousness into a cutout, in fact, its most original meaning is fusion. If you understand

i = AF + (1-a) b This fusion formula, you should understand what I'm doing. This is essentially the same as the second fusion principle.

In the narrow sense, fusion means matting.

For example, the famous Possion Fusion, right, if the moon map directly affixed to the sky, the rectangular edge is very obvious, the fusion of the left-hand image is very natural.

Code Visible http://blog.csdn.net/baimafujinji/article/details/46787837

In film technology, the matting method is often used to replace a person's scene. For example

12. Image Information Security

Mainly includes two content: 1) digital watermark (mainly used for multimedia copyright protection); 2) image encryption (mainly used for image information protection)

Example is the encryption effect I implemented with Magichouse (http://blog.csdn.net/baimafujinji/article/details/50500757)

Note that the area we are talking about is just the scope of image processing and does not involve machine vision. So there is no machine learning content, there is time for us to continue to discuss this aspect of things.

What is the study of digital image processing based on examples and codes?