Early detection of cancer through medical imaging has a critical impact on patient survival rates. There are many efforts for detecting early cancer in situ using modalities other than traditional medical optical imaging, which contain additional information over conventional micrographs of surface morphology acquired without staining. We analyzed the Mueller matrix components of human colon tissue obtained with an imaging polarimeter microscope at an illumination wavelength of 442 nm by principal components analysis in order to separate the traditional non-polarized gray image and to investigate the structure of the parameter space of polarization transformation by tissue. We also analyzed Mueller matrix by mapping it to a coherent matrix and performed eigenvalue analysis. The 1st to 4th principal components contain 99% of the information present in the images; polarization information contributes less than 10% of the information in the Mueller matrix. In one individual, 80% of the cancer was detected, without the first components which contains traditional non-polarized gray image for traditional diagnosis. Microscopic fine structures were observed, particularly in the 3rd and 4th principal components' score images. The entropy image of corrugated cancer tissue was smoother than that of the traditional gray image. There were several abnormal regions identified in adjacent regions of cancer, whose residues exceeded the noise level of the instrument used.