Research on feature extraction algorithms for computer vision is crucial. In some algorithms, the extraction of a high-complexity feature may solve the problem (for purposes such as target detection), but this will come at the expense of processing more data and requiring higher processing results. Color features do not require a lot of calculations. Simply convert the pixel values ​​in the digital image to reflect the values. Therefore, color features have become a better feature with their low complexity.
In image processing, we can analyze the color presented by a specific pixel by a variety of methods and extract its color feature components. For example, the color feature of a particular region is extracted by manually marking the region, and the region is represented by the average of the three components of a color space, or three color histograms can be established. Let us introduce the concept of color histogram and color moment.
Color histogram:
The color histogram is used to reflect the composition distribution of the image color, that is, the probability of occurrence of various colors. Swain and Ballard first proposed the method of image feature extraction using color histogram [40]. Firstly, the color histogram was obtained by stripping the three components of color space. Then, by observing experimental data, it was found that the image was rotated, scaled, The color histogram of the image after the fuzzy transformation does not change much, that is, the image histogram is not sensitive to the physical transformation of the image. Therefore, color features are often extracted and color histograms are applied to measure and compare the global differences of the two images. In addition, if the image can be divided into multiple regions and the foreground and background color distributions are significantly different, the color histogram exhibits a bimodal shape.
Color histograms also have their disadvantages: because the color histogram is the result of global color statistics, the positional features between pixels are lost. There may be several images with the same or similar color histograms, but their image pixel position distributions are completely different. Therefore, the relationship between the image and the color histogram is such that the color histogram does not achieve good results in identifying foreground objects.
Taking into account the above problems of the color histogram, the main tone histogram is generated. The so-called main tone histogram is based on the assumption that the values ​​of a few pixels can represent most of the pixels in the image, that is, the pixels with the highest frequency of occurrence are selected as the main color, and only the main color histogram composed of the main colors is used to describe one picture. image. Such a descriptor does not reduce the effect of matching by color features, because from a certain angle, pixels with a small frequency can be regarded as noise.
Color moment:
The color moment is an effective color feature, proposed by Stricker and Orengo [41], which uses the concept of moments in linear algebra to represent the color distribution in an image with its moment. The color distribution is described using a color first moment (average value), a color second moment (variance Variance), and a color third moment (skewness). Unlike color histograms, image description using color moments does not require quantization of image features. Since each pixel has three color channels of color space, the color moment of the image has 9 components to describe. Because of the small number of color moments, color moments are often combined with other image features.
Color set:
The above two methods are generally used for color comparison, matching, etc. between the two images globally or between regions, and the color set method is devoted to realizing the retrieval of large-scale images based on color realization. The method of color set is proposed by Smith and Chang [42]. After converting the color into the HSV color space, the image is divided into several regions according to its color information, and the color is divided into multiple bins. The color space is quantized to establish a color index, and then a binary image color index table is created. To speed up the search, you can also construct a binary search tree for feature retrieval.
Vaporesso Vape pen, Manufacture Vaporesso Vape, Wholesale Vaporesso Vapes
Shenzhen Xcool Vapor Technology Co.,Ltd , https://www.szxcoolvape.com