<p>The problem of the research is to find medical images of purity, high quality and free of impurities, which contributes to enabling doctors to obtain the results of analyzing the health status of each patient according to his disease data. Therefore, it was necessary to use discrete first chebysheve wavelets transform (DFCWT) technique in order to remove the associated impurities that appear in the medical images, and then analyze the results for all of the above, the algorithm DFCWT has been combined with and linking it to a neural network based on convolutional neural network (CNN) and this contributes to obtaining the results of analyzing image data with high accuracy and speed. The new algorithm proposed in this paper is based on deep learning finding the identification of kidney stones using DFCWT and the same process can be repeated on skin cancer, bones and fractures, processing by discrete first chebyshev wavelet transformation convolution neural network (DFCWTCNN).</p>

The rapid development of technology led to the need to keep pace with it, especially in the field of image processing, because of its importance in the need to get better results. In this paper, new discrete Chebyshev wavelet transforms (DChWT) derived from Chebyshev polynomials (ChP) were proposed and characterized. In terms of orthogonality and approximation (convergence) in the field of mathematics, (ChP) were qualified to transform into discrete wavelets called (DChWT), depending on the mother function with operators (s, r), and were used in image processing to analyze them in the low filter and the high filter so that the image is analyzed into coefficients. Proximity and detail coefficients, which lead to dividing the image into four parts, low left (LL), in which the proximity coefficients are concentrated while the rest of the parts are centered on the detail coefficients, which are high left (HL), low right (LR), and high right (HR), and image compression through the new filter, which has proven its efficiency at level (8) in our results. The results of the proposed wavelets in this work were reached in calculating quality standards in the image obtained after the compression process after comparing them with the results obtained using the standard wavelet used in HaarSymlet 2, Conflict 2, and Daubecheis 2. The most important of these standards is a mean square error (MSE), peak signal-to-noise ratio (PSNR), bit per pixel (BPP), compression ratio (CR), and table one. In this paper, the efficiency of the proposed new wavelets is explained after adding it to MATLAB and according to a specific program to drop in with the basic wavelets to take on their role in important tasks in the image processing area, like artificial intelligence

Classification is of great importance in the field of image processing, and convolutional neural networks (CNNs) have achieved great success in this field. Although CNN has proven to be a powerful technology for image recognition problems, it has failed in complex situations involving many realworld applications (for example, visual monitoring and automated driver assistance). Where it is difficult to detect a human in a series of images for various reasons. One of these reasons is the difference in the size of the human body, the height of the platform to which the camera is attached during the task of capturing accurate images, and the short training time in using the cameras, all of which are important factors to consider for the robustness and effectiveness of the human classification system. In this paper, a new deep CNN-based learning model is designed based on a new discrete waveform transformation (DWT) derived from discrete Hermit wavelet transform (DHWT) instead of modular wavelet, and the second stage is to train the convolutional neural network Hermit wavelets (HWCNN) is the most accurate and efficient deep learning.

Cars that violate the red light, and to increase the huge number of cars in violation, it is necessary to discover a system for identifying car plate numbers with the intervention of a computer, computer vision and neural networks segment and detail the number plates by designing regular algorithms to identify the number of license plates in violation. In this work, interest is in identifying the Iraqi car plate in order to know the place where the vehicle papers and the letters on which the vehicle depends and to know the location of the car were completed. The technique that was carried out in this work is to build new wavelets from polynomials by mathematical methods and discover a new algorithm using the MATLAB program to identify each number in the vehicle plate with a specific color by training a convolutional neural network (CNN) after analyzing the image using the new wavelets to identify the contents of the plate and good results have been reached. The accuracy level was reached with good values of up to 95%.