Research On Automatic Evaluation Of Photoshop Works Based On Image Similarity Comparison

Information technology has already became the mainstay of modern social development,image as an indispensable information carrier is widely used in various fields of information technology.In the meantime,there are various kinds of software with image editing function.Adobe's Photoshop software,with its powerful image processing capabilities,has became the most popular software.The curriculum of operational skills for Photoshop has been the important link in colleges and universities.And the examination method of this course is based on computer,which includes objective and subjective questions.Objective questions have been automatically reviewed by computer,but subjective operations are still dominated by manual evaluation.Existing automatic grading method for subjective questions adopted XML language which could interpretation the PSD source document information.This quantitative analysis method is tedious and not considering the actual aesthetic effect.In order to solve the problem of existing method,this paper starts with the automatic evaluation that point to the processing effect of examinee's actual image.By comparing the similarity between the examinee's actual image and the reference answer image,the work realized the purpose of automatic evaluation.The single feature similarity comparison of images had been carried out in the first step.Secondly,the comprehensive feature similarity comparison had been proposed.The main contents are as follows:1.The single feature such as color,shape and texture feature of image was extracted and the corresponding similarity had been calculated.Color components of image had been vector quantified by Gaussian mixture model.This method solves the problem that the traditional color histogram lacks the information of color spatial distribution relationship,and it makes a more comprehensive review of the color information.The optimal threshold segmentation method had been proposed which aimed at subject of selection operation in Photoshop software.The seven constant matrix of edge image were calculated as the shape characteristics,and the shape similarity of the image was reviewed.The article also calculated the image's roughness value and the contrast value,put forward the self-adaptive histogram and non-uniform quantization histogram,and the texture similarity of the image was properly reviewed.2.This paper put forward a comprehensive similarity evaluation method based on the fusion of single feature.By the weighted sum of the similarity of single feature,the reasonable evaluation result was obtained.The wavelet decomposition method was proposed to review image's overall similarity,it extracted the second wavelet decomposition figure of image's three-channel component and obtained the color and shape information,through this method the speed of the automatic review could be improved effectively.Review method based on region partition had been proposed,based on the image sub-block gradient,the image was divided into uniform area and the non-uniform area,respectively take different quantitative methods,got the review results which accord with human visual perception.3.A reasonable evaluation of the Photoshop works was proposed.By the method of comparing the histogram,this paper realized the reasonable score result in different image effect situations.Method in this paper can evaluate the answer image effectively.Compared with the current automatic evaluation method,the method of this paper is more suitable for the personalized examination at the present stage,and the review results are more consistent with the visual perception of human sight.