Underwater Image Quality Evaluation Database Based On Preference Label

The subjective image quality assessment database is of great significance for image retrieval and screening,evaluation and verification of image quality indicators,and learning objective image quality assessment models.The existing natural image quality assessment database is constructed by evaluating simulations image with different distortion types and same distortion type different degradation levels of the original image.Underwater images are different from natural images in the air,underwater images contain an interacting mixture of distortions due to the physical property of water and the instability of imaging systems,we cannot obtain original clear underwater images that can be used as a reference,the subjective image quality assessment method commonly used when establishing an existing image quality evaluation database is not suitable for underwater images.As far as we know,there is still no underwater image quality evaluation database for research reference,which further limits the development of underwater image objective quality evaluation methods.This article focuses on subjective underwater images quality assessment methods,the underwater image quality assessment database,and the objective underwater image assessment methods,the main research contents include:First,a preference label subjective underwater image quality assessment method is proposed.This method does not require the original image as a reference,turn pairwise ranking questions into a three choice questions,allowing observers to give "uncertainty" labels during the scoring process.We demonstrate that the obtained subjective image ranking correlates better with the human consciousness of quality difference for underwater images than the single stimuli image quality assessment.It is also accurate to reflect the gradual degradation in the underwater image sequence taken under control conditions.Second,based on the preference label subjective underwater image quality assessment method,the construction and expansion plan of the underwater image quality evaluation database is proposed.Designed an underwater image quality evaluation voting program of online progressive learning ranking,by combining laboratory evaluation and online evaluation,a adaptive extension underwater image quality ranking was established.This paper analyzes the important factors that affect the evaluation accuracy during online evaluation,and implements a 1,000 underwater image quality evaluations database with subjective evaluation values.Third,on the basis of the constructed underwater image quality evaluationdatabase,according to the scenes statistical characteristics related to image degradation and color-related statistical characteristics,using color histogram to classify underwater images according to the degenerated topics,and a multi-topics objective underwater image quality assessment method is designed.Experiments prove that the proposed method is superior to the existing objective underwater images quality evaluation methods and the quality evaluation model that adapts to the degradation theme can achieve more satisfactory results.