At present, microscopic observation of pathological sections is still the gold standard for pathological diagnosis of cancer. The complex production process and manual detection process of pathological sections make pathological detection subjective and inefficient. Polarization imaging technology is sensitive to sub wavelength structures, and exploring an objective and efficient method for identifying cancerous tissue using polarization images has unique advantages in enhancing pathological diagnostic capabilities.In this article, based on the Muller matrix measurement scheme of double wave plate rotation method, a backscatter polarization imaging system is built and upgrated equipped with a tunable zoom microscope lens to meet the requirements of different resolution fields. The slice data of unstained lung cancer and basal cell cancerare collected, and the Muller matrix is obtained from 30 polarization intensity images based on the Fourier coefficient relationship. In order to enhance the interpretability and strong correlation of the original data analysis paradigm in which the meaning of the Mueller matrix elements is unclear and the interpretation of the information for a single polarization parameter is limited, we report a polarization multi-parameter feature recognition and texture feature analysis method for cancerous tissue. To overcome the limitation that a single Muller matrix image cannot accurately and comprehensively identify the structure of pathological tissue, we introduce rotation invariants to obtain a high-dimensional polarization parameter set, then select regions of interest randomly and generate polarization multi-parameter feature curves to achieve multi-dimensional feature extraction and visualization of pathological regions, solving the problem of direct use of Muller matrix affected by direction. At the same time, in order to further obtain organizational information from derived parameters, 4 texture attributes from gray level co-occurrence matrix and 6 texture attributes from Tamura are calculated to assist in quantitative analysis.The proposed method is experimental verified, the following results are obtained: the characteristic curves of polarization parameter set obtained from 20 random samples in the normal and cancerous regions of lung cancer have a very high degree of overlap respectively, indicating that the polarization characteristics of the same type of tissue are generally similar, but appear to be significantly different in comparison, which perfectly conforms the previous analysis of a single parameter. The visualized polarization multi-parameter feature curve displays the complete polarization characteristics of normal and cancerous tissue of lung cancer in a very concise and clear manner, while also clearly showing the difference in curve trends between normal and cancerous tissue. This method is also applicable to basal cell carcinoma. The information distribution characteristics of normal and cancerous tissue of lung cancer are analyzed by fixing the texture dimension and polarization dimension respectively. When the texture feature is fixed, each polarization parameter has different degrees of discrimination effect. For example, when the contrast attribute is fixed, all the polarization parameters except for the parameter indicating linear polarization ability have good discrimination for lung cancer tissue and can be used as auxiliary tools for quantitative analysis; when the polarization dimension is fixed, the distribution of values for different textures on normal and cancerous tissue is different. For example, for the parameter indicating the angle of phase delay, the texture contrast, correlation, energy, and homogeneity of cancerous tissue are generally higher than those of normal tissue, and their corresponding six Tamura features have good discrimination, all of which have the potential to be used for quantitative analysis.According to the above research process and results, the fitted polarization multi-parameter feature curve restores the original high-dimensional polarization parameter set to two demensions, which can visually and efficiently identify the distribution of polarization differences between normal and cancerous tissue, and can intuitively obtain the information about the differences between different types of tissues in various polarization dimensions. At the same time, the results of texture analysis of lung cancer show that when the texture dimension is fixed, a single texture attribute can be a common quantitative indicator for multiple polarization images; when the polarization dimension is fixed, different texture attributes are expected to be multiple auxiliary quantitative indicators for a single polarization dimension. This method is fast and efficient, providing a new idea for auxiliary pathological detection and demonstrating good application prospects in clinical practice.