Color Image Processing

The use of color is important in image processing because:

 Color is a powerful descriptor that simplifies object identification and extraction.
 Humans can discern thousands of color shades and intensities, compared to about only
two dozen shades of gray.

Color image processing is divided into two major areas:

 Full-color processing: images are acquired with a full-color sensor, such as a color TV
camera or color scanner.
 Pseudocolor processing: The problem is one of assigning a color to a particular
monochrome intensity or range of intensities.

Color Fundamentals
Colors are seen as variable combinations of the primary color s of light: red (R), green (G), and
blue (B). The primary colors can be mixed to produce the secondary colors: magenta (red+blue),
cyan (green+blue), and yellow (red+green). Mixing the three primaries, or a secondary with
its opposite primary color, produces white light.

RGB colors are used for color TV, monitors, and video cameras.
However, the primary colors of pigments are cyan (C), magenta (M), and yellow (Y), and the
secondary colors are red, green, and blue. A proper combination of the three pigment primaries,
or a secondary with its opposite primary, produces black.
CMY colors are used for color printing.

Color characteristics
The characteristics used to distinguish one color from another are:
 Brightness: means the amount of intensity (i.e. color level).
 Hue: represents dominant color as perceived by an observer.
 Saturation : refers to the amount of white light mixed with a hue.

Color Models
The purpose of a color model is to facilitate the specification of colors in some standard way. A
color model is a specification of a coordinate system and a subspace within that system where
each color is represented by a single point. Color models most commonly used in image
processing are:

RGB model for color monitors and video cameras

 CMY and CMYK (cyan, magenta, yellow, black) models for color printing
 HSI (hue, saturation, intensity) model

The RGB color model

In this model, each color appears in its primary colors red, green, and blue. This model is based
on a Cartesian coordinate system. The color subspace is the cube shown in the figure below. The
different colors in this model are points on or inside the cube, and are defined by vectors
extending from the origin.
All color values R, G, and B have been normalized in the range [0, 1].
However, we can represent each of R, G, and B from 0 to 255.Each RGB color image consists of
three component images, one for each primary color as shown in the figure below. These three
images are combined on the screen to produce a color image. The total number of bits used to
represent each pixel in RGB image is called pixel depth. For example, in an RGB image if each
of the red, green, and blue images is an 8-bit image, the pixel depth of the RGB image is 24-bits.

The CMY and CMYK color model

Cyan, magenta, and yellow are the primary colors of pigments. Most printing devices such as
color printers and copiers require CMY data input or perform an RGB to CMY conversion
internally. In printing, combining equal amounts of cyan, magenta, and yellow produce muddy-
looking black. In order to produce true black, a fourth color, black, is added, giving rise to the
CMYK color model.
The HSI color model
The RGB and CMY color models are not suited for describing colors in terms of human
interpretation. When we view a color object, we describe it by its hue, saturation, and brightness
(intensity). Hence the HSI color model has been presented. The HSI model decouples the
intensity component from the color-carrying information (hue and saturation) in a color image.
As a result, this model is an ideal tool for developing color image processing algorithms.
The hue, saturation, and intensity values can be obtained from the RGB color cube. That is, we
can convert any RGB point to a corresponding point is the HSI color model by working out the
geometrical formulas.
Pseudo Color Image Processing
Basics of full color Image Processing

Colour images contain more information than grayscale images because each pixel has multiple
values (usually Red, Green, and Blue components). Processing such images requires specialized
techniques.

Two Main Approaches in Colour Image Processing

1. Component-wise Processing (Channel-by-Channel)

 Each colour image has three separate grayscale images (R, G, and B channels).
 This approach treats each channel independently and processes them separately.

Steps:

1. Split the image into its three components: R, G, and B.

2. Apply grayscale image processing techniques (like filtering, histogram equalization) to
each channel separately.
3. Merge the channels back to form the processed colour image.
 Example:

 Apply edge detection separately on R, G, and B channels, then combine the results.

2. Vector-wise Processing (Pixel as a Vector)

 Here, each pixel is considered a 3D vector (R, G, B).

 Processing involves vector operations, treating the colour pixel holistically.

Techniques:

 Vector filtering (e.g., vector median filter),

 Vector quantization (used in compression),
 Colour clustering (e.g., k-means in RGB space).

Advantages:

 Maintains the relationship between channels,

 More accurate in color-based segmentation and recognition,
 Reduces risk of color distortion.

COLOR TRANSFORMATION

Color transformation refers to the conversion of an image from one color space to another in
order to simplify or enhance processing, analysis, or display of color information.

Each color space has a specific use depending on the application (e.g., RGB for display, HSV for
segmentation, YCbCr for compression).

Image Segmentation: Active Contours: Snakes and Level Sets:

Active contours, commonly referred to as snakes, and level sets are two related techniques used
in image segmentation. They both aim to automatically detect and delineate object boundaries
within images. Here's an overview of each technique:

Active Contours (Snakes):

Principle: Active contours, or snakes, are parametric curves that evolve within an image to
delineate object boundaries. They are attracted to features like edges or intensity gradients while
resisting factors like noise and weak edges.
Algorithm:
1. Initialization:
 Start with an initial contour, often a simple shape like a circle or polygon, placed
close to the object boundary.
2. Energy Minimization:
 Minimize an energy functional that defines the snake's behavior.
 The energy functional consists of:
  Internal energy: Encourages smoothness and regularity of the contour.
 External energy: Drives the contour towards object boundaries.
 The snake evolves by iteratively minimizing the combined internal and external
energies.
3. Evolution:
 At each iteration, the snake deforms according to internal and external forces, moving
towards object boundaries.
 Typically, this is done by solving a set of differential equations, where the snake's
shape evolves over time.
4. Convergence:
 The process continues until the snake stabilizes, converging to the object boundaries.
Advantages:
Effective in segmenting objects with well-defined boundaries.
 Can handle complex shapes and topologies.
Disadvantages:
 Sensitive to initial contour placement and parameters.
 Struggles with concavities, noise, and weak boundaries.

Level Sets:
Principle: Level set methods represent object boundaries implicitly as the zero level set of a
higher-dimensional function. These methods allow for topological changes in the evolving
contours.
Algorithm:
1. Initialization:
 Start with an initial level set function, which represents the initial contour.
2. Propagation:
 Evolve the level set function over time using partial differential equations (PDEs).
 The PDEs drive the zero level set towards object boundaries while maintaining
smoothness.
3. Evolution:
 The level set function evolves by solving the PDEs, causing the zero level set to
propagate and deform.
4. Convergence:
 The process continues until the level set stabilizes, converging to the object boundaries.
Advantages:
 Handles topological changes naturally, such as object splitting and merging.
 Less sensitive to parameter initialization compared to snakes.
Disadvantages:
 Computationally intensive, especially for 3D images.
 Can be slower than snakes due to numerical computations involved.
Applications:
 Medical Imaging: Segmenting organs, tumors, and structures from MRI, CT, or
ultrasound images.
 Biomedical Analysis: Identifying cells, nuclei, or structures in microscopy images.
 Object Tracking: Tracking moving objects in videos or sequences of images.
  Computer Vision: Extracting objects from natural scenes or detecting boundaries for
recognition tasks.

Thresholding in Segmentation:
Thresholding is one of the simplest and most widely used techniques in image segmentation. It
involves dividing an image into foreground (object) and background regions based on pixel
intensity values.

Image and Threshold:

 Given an input grayscale image, the goal is to partition it into two regions: foreground
(objects of interest) and background.
 Thresholding requires selecting a threshold value, which acts as a cutoff point to separate
pixels into these two regions.

2. Global Thresholding:
 In global thresholding, a single threshold value is applied to the entire image.
 Pixels with intensities greater than or equal to the threshold value are classified as
foreground (usually assigned a value of 1 or 255), while pixels with intensities below the
threshold are classified as background (usually assigned a value of 0).
 The threshold value can be manually chosen based on domain knowledge or
automatically determined using techniques like Otsu's method, which maximizes the
separability of the foreground and background intensity distributions.

Global Thresholding using Otsu’s Method:

Global thresholding using Otsu's method is a technique to automatically determine an optimal

threshold value for segmenting an image into foreground and background regions. It works by
maximizing the between-class variance of pixel intensities, effectively finding the threshold that
best separates the two classes. Here's how it works:
1. Histogram Calculation:
 First, calculate the histogram of the input grayscale image. The histogram represents the
distribution of pixel intensities in the image.
2. Normalization:
 Normalize the histogram such that the sum of all histogram values equals 1. This step
ensures that the histogram represents the probability distribution of pixel intensities.
3. Cumulative Distribution Function (CDF):
 Compute the cumulative distribution function (CDF) from the normalized histogram.
The CDF represents the cumulative probability of each intensity value.
4. Mean Intensity:
 Compute the mean intensity of the image, denoted as total μtotal, using the formula:
μtotal=Σi=0L−1i⋅p(i) where L is the number of intensity levels (typically 256 for an 8-
bit image) and p(i) is the normalized histogram value at intensity level i.

Otsu's method provides an automatic way to determine an optimal threshold value without the
need for manual intervention. It's particularly useful when there's a clear separation between
foreground and background intensities in the image. However, it may not perform well in cases
of uneven illumination or complex backgrounds.

Multiple Thresholding:

Multiple thresholding is a segmentation technique where an image is divided into multiple

regions using more than one threshold value. This technique is useful when the image contains
multiple objects or regions with distinct intensity levels, and simple binary segmentation (using a
single threshold) is insufficient. Here's how multiple thresholding
works:
1. Histogram Calculation:
Compute the histogram of the input grayscale image, representing the distribution of pixel
intensities.
2. Threshold Selection:
Determine the number of thresholds n needed based on the characteristics of the image and the
desired segmentation. Select n−1 threshold values to divide the image into n segments.
3. Thresholding:
Apply the threshold values to the image to segment it into n regions. Pixels with intensities
falling within each threshold range are assigned to the
corresponding region.
4. Segmentation:
 After thresholding, the image is divided into n segments, each representing a different
region with similar intensity characteristics.
 The resulting segmented regions can be further processed or analyzed individually.
5. Visualization:
 Each segmented region can be displayed with a unique grayscale intensity or color to
distinguish it from other regions.
 For example, each segmented region might be assigned a different color to visualize the
segmentation.

Multiple thresholding offers a flexible approach to segmenting images with complex intensity
distributions, allowing for the identification of multiple regions with different characteristics. By
dividing the image into multiple segments, it provides finer detail and granularity in
segmentation results compared to simple binary thresholding.

Variable Thresholding:
Variable thresholding, also known as adaptive thresholding, is a segmentation technique where
the threshold value is determined locally for each pixel in the image, based on its neighborhood.
This method is particularly useful for images with varying illumination or background intensity.
Here's how variable thresholding works:
1. Divide Image into Patches:
 The input image is divided into smaller patches or windows of fixed size.
2. Calculate Threshold for Each Patch:
 For each patch, compute a threshold value that is adapted to the local characteristics
of that patch.
  Common methods for calculating the local threshold include:
Mean Thresholding: Compute the mean intensity of the pixel values within the patch and use it
as the threshold.
Gaussian Thresholding: Compute the weighted average of pixel intensities using a Gaussian-
weighted kernel.
Median Thresholding: Use the median intensity value of the pixel values within the patch as
the threshold.

Apply Thresholding Apply the computed local threshold to each pixel in the corresponding
patch.
 Pixels with intensities greater than or equal to the local threshold are classified as
foreground, while pixels below the threshold are classified as background.
Applications:
 Document Processing: Adaptive thresholding is commonly used in OCR (Optical
Character Recognition) and document analysis to segment text from the background,
especially in scanned documents with uneven lighting.
 Biomedical Imaging: In medical imaging, adaptive thresholding can help segment
structures like bones or organs from X-ray images with varying intensity.
Challenges:
 Parameter Selection: Choosing the appropriate size of the patch and the method for
calculating the local threshold can be critical and may depend on the characteristics of
the image.
 Computational Cost: Adaptive thresholding methods can be computationally
expensive, especially for large images or when using complex thresholding techniques.
.
Segmentation by Region Growing, Splitting and Merging:
Segmentation by region growing, splitting, and merging are techniques used to partition an
image into regions or segments based on pixel similarity. These methods iteratively group pixels
into regions that exhibit similar characteristics, such as intensity, color, or texture.
Here's an explanation of each technique:
1. Region Growing:
Principle: Region growing starts with a set of seed points or seeds and iteratively adds
neighboring pixels to the region if they satisfy certain similarity criteria.
Algorithm:
1. Select seed points either manually or automatically based on predefined
criteria.
2. Initialize an empty region for each seed point.
3. For each seed point, check neighboring pixels.
4. If the neighboring pixel satisfies the similarity criteria (e.g., intensity
difference, color similarity), add it to the corresponding region.
5. Repeat step 4 for all neighboring pixels until no more pixels can be added.
Termination Condition: Typically, region growing stops when no more pixels meet the
similarity criteria or when the region reaches a predefined size.

Advantages:
 Simple to implement.
  Works well for regions with uniform characteristics.
Disadvantages:
 Sensitive to seed selection and initial conditions.
2. Region Splitting:
Principle: Region splitting starts with the entire image as one region and iteratively splits
regions into smaller ones based on certain criteria until the segmentation is satisfactory.
Algorithm:
1. Start with the entire image as one region.
2. Evaluate each region to determine if it should be split based on predefined
splitting criteria (e.g., variance, texture complexity).
3. If a region meets the splitting criteria, divide it into smaller regions.
4. Repeat steps 2-3 for each region until no more regions meet the splitting
criteria.
Termination Condition: Region splitting stops when no more regions meet the splitting criteria.
Advantages:
 Can handle images with varying characteristics.
 Less sensitive to initial conditions compared to region growing.
Disadvantages:
 May lead to over-segmentation if splitting criteria are too relaxed.
 Computationally expensive for large images.
3. Region Merging:

Principle: Region merging starts with each pixel as a separate region and iteratively merges
neighboring regions based on certain criteria until the segmentation is satisfactory.

Algorithm:
1. Start with each pixel as a separate region.
2. Evaluate pairs of neighboring regions to determine if they should be merged
based on predefined merging criteria (e.g., similarity in intensity, color,
texture).
3. If a pair of regions meets the merging criteria, merge them into a single region.
4. Repeat steps 2-3 until no more regions meet the merging criteria.

Termination Condition: Region merging stops when no more regions meet the merging criteria.

Advantages:
 Can handle over-segmented images.
 Less sensitive to initial conditions compared to region growing.

Applications:
 Medical Imaging: Segmenting organs or tissues from medical images like MRI or CT
scans.
 Satellite and Aerial Imaging: Identifying land cover types, urban areas, or vegetation
from satellite images.