Xử lý ảnh || Chương 2. Cải thiện ảnh || Các phép toán trên điểm ảnh

Overview

The video introduces spatial-domain image enhancement, which changes individual pixel values to make an image more useful for a particular purpose. Enhancement can reveal details, improve contrast, or convert a grayscale image into a binary one. What counts as “better” depends on the application.

Image Enhancement Approaches

The instructor divides image-enhancement methods into two broad groups:

• Spatial-domain methods operate directly on image pixels. They include point operations, spatial filtering, and histogram-based methods.
• Frequency-domain methods enhance an image by manipulating its frequency components. These are mentioned but left for later lessons.

For point processing, the general relationship is s = T(r), where r is the input pixel value, T is the operation applied, and s is the resulting pixel value.

Point Operations Covered

Image Inversion (Negative)

Image inversion reverses grayscale brightness values using:

s = L − 1 − r

Here, L is the number of possible gray levels. For a common 8-bit grayscale image, L = 256, so the operation becomes s = 255 − r. Inversion can make light or gray details in dark regions easier to see.

Logarithmic Transformation

The logarithmic transformation uses:

s = c log(1 + r)

Here, c is a constant. The transformation changes the image’s intensity distribution and can make some details more visible.

Gamma (Power-Law) Transformation

The gamma transformation uses:

s = c r^γ

Here, c and γ are positive constants. Changing these parameters changes the result’s appearance. The instructor demonstrates comparing outputs produced with different gamma values.

Thresholding

Thresholding converts a grayscale image into a binary image with two possible pixel values:

• If a pixel is below the threshold, set it to 0.
• Otherwise, set it to the maximum value.

The video demonstrates a threshold value of 117 and compares the original grayscale image with the thresholded result.

Coding Demonstration

The instructor demonstrates implementing the operations and displaying original and processed images side by side using Python, OpenCV, and Matplotlib. The walkthrough broadly follows these steps:

1. Install the required libraries.
2. Load an image.
3. Define a function for the desired pixel transformation.
4. Apply the function to the image.
5. Display and compare the original and processed results, using titles and multiple panels where useful.

The examples include separate displays for image inversion, logarithmic transformation, several gamma settings, and thresholding. The instructor says the code is linked in the video introduction.

Speakers and Sources

• Speaker: One instructor or presenter, unnamed in the subtitles.
• Technical tools mentioned: OpenCV and Matplotlib.
• No other speaker or clearly identifiable source is established by the subtitles.