Visual Perception of Images

Transcription

1 Visual Perception of Images A processed image is usually intended to be viewed by a human observer. An understanding of how humans perceive visual stimuli the human visual system (HVS) is crucial to the design of many image processing algorithms. We need to understand: photometric properties of physical world spatial response of the HVS temporal response of the HVS spectral/color response of the HVS Stanley J. Reeves ELEC Digital Image Processing 1 / 16

2 Structure of Human Eye cornea - hard, fixed lens lens - soft, flexible lens that allows the focus of the eye to change retina - the inner membrane on which the image is focused and sensed Choroid Sclera Posterior chamber Retina Optic disc Optic nerve Zonular fibres Iris Vitreous humour Pupil Lens Fovea Cornea Anterior chamber (aqueous humour) Hyaloid canal Retinal blood vessels Ciliary muscle Suspensory ligament Stanley J. Reeves ELEC Digital Image Processing 2 / 16

3 Visual System Basics rods - a class of photoreceptor cells that do not distinguish wavelength but are sensitive to dim light (scotopic or dim-light vision) cones - a class of photoreceptor cells that are sensitive to wavelength at high light levels (photopic or bright light vision) three types of cones, roughly corresponding to bandpass responses around red, green and blue wavelengths (long, medium, and short) the eye perceives color as a weighted combination of these three types of cones Stanley J. Reeves ELEC Digital Image Processing 3 / 16

4 Brightness Adaptation The HVS can adapt to distinguish up to light intensity levels from the scotopic threshold to glare level. The intensity range discriminated in one instant (the brightness adaptation level) is rather small. Light sensitivity increases in dim light and decreases in bright light to accommodate a wide range of illumination levels. Stanley J. Reeves ELEC Digital Image Processing 4 / 16

5 Weber s Law Weber s Law: A just-noticeable difference (JND), denoted I, in a stimulus is proportional to the magnitude of the original stimulus I. Contrast sensitivity = I/ I. At very low luminance, detector noise and ambient light tend to reduce sensitivity, so the stimulus appears black. At very high luminance, the very bright background tends to saturate detector sensitivity, thereby reducing sensitivity by blinding the subject. Stanley J. Reeves ELEC Digital Image Processing 5 / 16

6 Simultaneous Contrast The brightness perceived by our visual system is a function of both the illumination emitted by the object and also its surroundings. Stanley J. Reeves ELEC Digital Image Processing 6 / 16

7 Simultaneous Contrast The brightness perceived by our visual system is a function of both the illumination emitted by the object and also its surroundings. Although the central patch is exactly the same in each case, simultaneous brightness contrast makes it appear to vary from dark to light as the background is changed from light to dark. Stanley J. Reeves ELEC Digital Image Processing 6 / 16

8 Mach Band Effect The luminance of the above squares increases stepwise. In each strip the right edge looks darker than left edge. This is due to lateral inhibition. The response at a given receptor is inhibited by the response of neighboring receptors. Stanley J. Reeves ELEC Digital Image Processing 7 / 16

9 Spatial Response of the HVS 1 0 Spatial step response of the HVS 1 0 Spatial impulse response of the HVS Stanley J. Reeves ELEC Digital Image Processing 8 / 16

10 Contrast Sensitivy Function (CSF) demo Contrast sensitivity functions of seven age groups (after Schieber, 1992) Stanley J. Reeves ELEC Digital Image Processing 9 / 16

11 Spatial-Frequency Effects Jekyll and Hyde Face Example Stanley J. Reeves ELEC Digital Image Processing 10 / 16

12 Brightness Constancy We compensate psychologically for lighting effects. This ability allows us to see objects in terms of inherent reflectance rather than the actual amount of light reflected. Called brightness constancy. A white paper appears white to us even when it is in shadow or in a relatively dark room. Cameras must make exposure corrections indoors; otherwise, the picture would not look the same as we see with our eyes. Stanley J. Reeves ELEC Digital Image Processing 11 / 16

13 Shadow Effects Effect of brightness adaptation and brightness constancy This photo shows actual light intensity reflected from the location. This photo shows how we interpret object brightness based on prior knowledge of lighting effects and not simply the intensit of light reflected. Stanley J. Reeves ELEC Digital Image Processing 12 / 16

14 Shadow Effects Another Example Stanley J. Reeves ELEC Digital Image Processing 13 / 16

15 Color Adaptation Our visual system adapts its color sensitivity in a manner that tends to make the illumination look colorless. Color film can achieve appropriate color balance only when exposed under specific illumination for which the film is designed. Stare at a point on the picture for 30 s. Then look at a blank wall. You will see an after-image due to local color adaptation. Stanley J. Reeves ELEC Digital Image Processing 14 / 16

16 Noise Masking Noise in an image is less visible in regions of the image that are busy but more visible in smooth, flat regions. We can see that noise is noticeable in the smooth parts of the picture (clouds and sidewalk) and less noticeable in the busy parts (bushes and road). Stanley J. Reeves ELEC Digital Image Processing 15 / 16

17 Bloch s Law of Temporal Summation Flashes of light will appear to be the same if they have equal total energy and last less than approximately 30 ms in moderate illumination. Flashes of light at about 50 to 60 Hz cannot be distinguished from a constant light with the same average energy. This frequency is known as the critical fusion frequency (CFF). The flickering effect can be noticed when we see a TV monitor at an angle. CFF is higher for larger objects than smaller. These effects have to be considered while coding or rendering video, as blurring will appear if motion takes place within a 20 ms interval. Stanley J. Reeves ELEC Digital Image Processing 16 / 16