Fundamentals of Computer Vision and its applications

The Computer will visualize and identify information in the form of images, process the information in the place of a human. It will be programmed to do a specific task in image processing. From machine learning to advanced deep learning models, will enable machines to identify the information and to work for the purpose. Computer vision seems to be a high level process and it is used for image recognition to image analysis. It will give more interpretation of images based on the features identified.

Figure 1: View on Computer vision and Human Vision (Source: https://algorithmxlab.com/blog/computer-vision)

Interesting and important Applications of Computer Vision

• CV in Medical field
  • To diagnose disease
  • To identify the stages of disease affected
  • To diagnose the general health conditions of patients, etc.,
• CV in remote sensing
  • To identify the features in a satellite image
  • To identify different areas like land, water and greeneries in the satellite image
  • To analyze the satellite images before and after natural disasters
  • To analyze the satellite images for agricultural related issues, etc.,
• CV in Automation Industries
  • To detect faults on manufactured devices
  • To detect the performance of the system developed
  • To implement automation on production mechanism
  • Autonomous vehicles, etc.,

Computer Vision in the field of Medicine

The application fields and purpose specified above are not limited. In this blog brief detail of Computer Vision on Medicine has been elaborated. Advantages of using computer vision in healthcare industries are briefed below.

• Accurate diagnosis of disease will be obtained through Computer vision, and it has been widely used to suggest an exact identification of the syndrome. This will be very much useful in many cases such as for cancer cell detection, tumor detection, etc.,
• Detection of abnormalities at an earlier stage some of the disease will not be giving physical symptoms at earlier stage, but by analyzing the scan images we can detect the stage of affected area earlier and we can react and take medications accordingly. It will decrease the risk factor of human life.
• Computer vision in the medical field will assist Doctors in diagnosis of disease and can save time for earlier detection and quick processing. It will also create awareness on patients about their health and medication.

In Medical Image processing information are extracted from the patient’s medical images to diagnose the disease. Various algorithms are available to extract the features from the scan images like tumor detection, blood flow detection, fibroid identification, cancer cell detection, brain structure analysis, arteriosclerosis, the other harmful disease can be analyzed using computer vision technologies.

Image Enhancement

Applications of computer vision in the field of medical domain start from image enhancement. Images taken from ultrasonic scan or X-ray or MRI or CT scan needs to be enhanced to get an accurate analysis of disease. This is due to influence of noise present during the scanning process.

In general image enhancement can be broadly classified into two categories.

• Spatial domain image enhancement
• Frequency domain image enhancement

Spatial domain technique, directly deal with the pixel values present in the image. Enhancement of the image is done by manipulating the pixel values in the original image. It can be done through filtering operations like mean filtering, median filtering, min max filtering, etc.,

Frequency domain method is enhancing image after converting it using any of the image transform techniques, the input medical image is transformed into frequency domain image using Fourier transform, Wavelet transform, Haar transform, Hadamard transform etc., after enhancing finally the original image is obtained by taking inverse transform of an image. These process of enhancement are performed to change the brightness and contrast of the image to have a better perception to analyze the disease accurately.

Feature Extraction

Feature extraction of one of the important phase of medical image processing, based on the identified feature the diagnosis will be done by machine learning algorithms. Feature extraction methodologies will analyze objects and images to extract the most important features which represent the exact details of region of interest. Some of the features that can be identified for medical image are intensity histogram features and Gray Level Co-Occurrence Matrix (GLCM) features.

Intensity Histogram features like mean, standard deviation, entropy, variance, energy, skewness, etc.,. can be identified in the images. Based on these features we can classify the images as normal or abnormal image and further processing can be derived to find the level of abnormality in the images.

GLCM features are a statistical analysis method which identifies the pixel relations in spatial domain. Some of the features are Cluster Prominence, Autocorrelation, Contrast, Correlation, Energy variations, average, variance, Total entropy, Homogeneity, Maximum probability, Sum of squares, Difference variance, Difference entropy, Information measure of correlation, information measure of correlation, Inverse difference normalized, etc., In this method the spatial relationship is identified based on the pixel of interest and the pixel which is horizontally adjacent. Each pixel element (i, j) in the GLCM feature analysis  is the sum of the number of times that the pixel with the value i occurred in the specified spatial relationship to a pixel with value j in the input image.

Feature Selection

Feature selection will be used to improve the calculation accurateness and reduces the execution time.  The above said performance is attained by eliminating inappropriate, repetitive and unwanted features.  It absorbs the part of features, which is used to achieve the best performance in terms of accuracy and computation time. It performs the Dimensionality reduction. Features are identified by using search algorithms. Some of the feature selection algorithms are Sequential forward Selection, Sequential Backward selection, Principle component Analysis and Particle Swarm Optimization, etc,. Sometimes the hybrid method of combining two or more algorithms will work better and improve the performance in an efficient way for optimal feature selection. After feature selection, diagnosis of medical images will be done using machine learning algorithms with computer vision technology by comparing the normal and abnormal images.

Conclusion

In this blog a preface to Computer vision and its application in the medical field is briefly explained. The process of disease diagnosis is done by image enhancement, feature extraction, feature selection. Further the image processing can be done with improvised algorithms for