Computer Vision – Fundamentals of Human Vision Hanyang University Jong-Il Park Introduction Understanding of Mechanism of Human Vision To construct the measures of image fidelity & intelligibility To design and evaluate image processing algorithms and imaging systems Division of Electrical and Computer Engineering, Hanyang University Brightness (Perceived Luminance) Light ~ radiant energy which, by its action on the organs of visions, enables them to perform their function of sight Spectral energy distribution of the light source L( λ ), λ = 350nm ~ 780nm Light received from an object I ( λ ) ρ( λ )L(λ ) ρ( λ ) : reflectivity or transmissivity of the object Division of Electrical and Computer Engineering, Hanyang University Human Eye visible range : 350 nm < wavelength < 780 nm photoreceptors of the retina rods : about 75-150 millions cones : about 6.5 millions (Color Vision) scotopic vision : rods (dark environment) mesopic vision : rods + cones (middle range) photopic vision : cones (bright environment) rods are more sensitive to light than the cones Division of Electrical and Computer Engineering, Hanyang University Distribution of Photoreceptors Division of Electrical and Computer Engineering, Hanyang University Division of Electrical and Computer Engineering, Hanyang University The Human Visual System 공막 홍채 (눈알의) 맥락막 각막 (안구의) 수양액 Shape of lens, rather than the distance between lens and screen, is changed Division of Electrical and Computer Engineering, Hanyang University The Human Visual System (cont.) Division of Electrical and Computer Engineering, Hanyang University Eye Physiology Luminance (or Intensity) : independent of luminance of the surrounding object Relative luminous efficiency function f ( x, y) I ( x, y, )V ( )d Light distribution Luminosity Function (Relative Luminous Efficiency Function) Division of Electrical and Computer Engineering, Hanyang University Brightness Contrast Brightness : dependent upon the surroundings Division of Electrical and Computer Engineering, Hanyang University Brightness adaptation Dynamic range ~ 1010 Division of Electrical and Computer Engineering, Hanyang University Intensity Discrimination Experiment Division of Electrical and Computer Engineering, Hanyang University Weber’s Law Division of Electrical and Computer Engineering, Hanyang University (100,101) (100,102) (100,105) (150,151) (150,153) (150,162) Division of Electrical and Computer Engineering, Hanyang University I C 0.02weber fraction I I For small value of : small changein intensityis discriminable I good brightnessdiscrimination rod cone typicalweber ratioas a fuctionof intensity Division of Electrical and Computer Engineering, Hanyang University Brightness(cont.) Mach Bands Division of Electrical and Computer Engineering, Hanyang University MTF of the Visual System Measurement of visual system in frequency domain MTF: Modulation Transfer function H (1 , 2 ) H ( ) A exp 0 0 12 22 cycles/degree Useful Values for ImageCoding Applications A 2.6, 0.0192, 1.1 0 (0.114) 1 8.772, . peak frequency: 8 cycles/degree Division of Electrical and Computer Engineering, Hanyang University Isopreference k: # of bits/pel gray-level resolution N: # of pels spatial resolution Preference depends on image! Division of Electrical and Computer Engineering, Hanyang University Optical illusions Division of Electrical and Computer Engineering, Hanyang University Moire pattern Division of Electrical and Computer Engineering, Hanyang University Image Fidelity Criteria Goal : Image quality measurement Performance evaluation of image processing techniques or systems Quantitative Criteria 1 2 Mean square criterion : ms MN SNR(signal-to-noise ratio) : PSNR(peak-to-peak SNR) : M N u(m, n) u(m, n) 2 m 1 n 1 2 SNR 10log10 2 ms PSNR 10log10 (peak - to - peak value) 2 2 ms Division of Electrical and Computer Engineering, Hanyang University Image Fidelity Criteria (cont.) Subjective Criteria Division of Electrical and Computer Engineering, Hanyang University Perception of Intermittent Light Consider a light that flashes on for a brief duration N times/sec Perception depends on its frequency ( N cycles/sec) small N : Flashes appear separated in time increase N : unsteady flicker, unpleasant increase N further : Continuous light perception Fusion frequency : Frequency at which an observer begins perceiving light flashes as continuous light Critical Fusion frequency (CFF) : about 50 ~ 60 Hz. Division of Electrical and Computer Engineering, Hanyang University Perception of Intermittent Light (cont.) Higher fusion frequency for larger size and larger intensity of the flickering object very dim, small light : A few cycles/sec very bright, large light : Over 100 cycles/sec Examples of intermittent light fluorescent light : Over 100 times/sec motion picture : 24 frames/sec with 1 frame shown twice TV monitor : 30 frames/sec, 2fields/frame 60 fields/sec (NTSC system) Division of Electrical and Computer Engineering, Hanyang University Empirical Observation Sharper images look better Same noise in uniform background region is more visible than noise in edge areas (spatial masking) Same noise in dark areas is more visible than noise in bright areas Same amount of artificial noise appears worse than natural looking noise Division of Electrical and Computer Engineering, Hanyang University Colorimetry The perceptual attributes of color (HIS system) Intensity : the amount of light, perceived luminance ex) distinction between dark grey and light grey hue : the color as described by wave length ex) distinction between red and yellow saturation : the amount of color that is present ex) distinction between red and pink the vividness of color Three primaries : Red, Green, Blue (RGB) Division of Electrical and Computer Engineering, Hanyang University Division of Electrical and Computer Engineering, Hanyang University Color representation Hue varies along the circumference Saturation varies along the radial distance } Chromaticity Division of Electrical and Computer Engineering, Hanyang University Eg. Color representation Division of Electrical and Computer Engineering, Hanyang University Color absorption spectra T hereare threedifferenttypesof conesin theretina with absorptionspectraSi , i 1,2,3 i (C) : spectral response T hus,coloredlight,C will produce a colorsensation,given by i (C) max min Si ( )C( )d , i 1,2,3 Division of Electrical and Computer Engineering, Hanyang University Color Vision Model e1 C λ S1dλ e2 C λ S 2 dλ e3 C λ S3 dλ, Si λ : spect ralsensit ivity d1 loge1 d 2 loge2 loge1 d 3 loge3 loge1 luminance chromaticity of a color Details to be covered later Division of Electrical and Computer Engineering, Hanyang University