Color and the Human Eye

Transcription

1 Color Images

2 Source: wikipedia Biophysical basis: Color and the Human Eye - The human eye contains 3 kinds photoreceptive cells (cone) cells that respond to different wavelengths: L : longer wavelengths M : medium wavelengths S : shorter wavelengths

3 Source: wikipedia Color and the Human Eye When light hits the cones, an electrical signal is transmitted to the brain. The relative intensity of the signal from the different types of cone cells is then mapped by the brain into what we interpret at color. Example: Which cones would detect light emitted at 5500 Angstroms? Which type of cone cell has the greatest response at this wavelength?

4 Source: wikipedia Color and the Human Eye What this means is that the eye effectively only sees three different primary colors. All other colors can be obtained by varying the relative intensities of these colors.

5 Color and The Non-Human Eye Not all types of animals have the same number of types of cones. Humans: 3 types (color blindness normally due to 1 or more type not working) Birds: 4 types (UV channel) Bees: 3 types (see UV, but not red) Most mammals (e.g. dogs): 2 types From Dr. M. Nofziger s notes on optics at the University of Arizona Source: wikipedia

6 What Humans can see What dogs and cats can see What bulls can see

7 Source: wikipedia Making a Color Image The net result is that to make a color image you need 3 images taken with different filters. - More doesn t help since the eye can only see three. - A true color image is one that uses three filters that roughly correspond to the transmission regions for the different types of cones. - Johnson BVR filters are reasonably matched to S,M,L cones (also close to BGR from CTO filter set)

8 Source: wikipedia Below: red lines are wavelengths of peak throughput for BVR filters Above: red lines are wavelengths of peak response for eye

9 False Color There s no fundamental reason you have to map information in true color. You can easily map and wavelength images to red, green, blue.

10 Kitakami River, Japan Kitakami River From Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) March 14, 2011 Jan 16, 2011

11 Red: F658N(NII), Green: F656N(Halpha), Blue: F502N(OIII) NGC

12 RGB Color Image RGB (which stands for Red-Green-Blue) is the most common format for storing color images. - For each color, the image is renormalized to have integer values running from 0-255, where 0 is dark and 255 is maximum brightness > 16.6 million yellow(255,255,0) green(0,255,0) cyan(0,255,255) red(255,0,0) blue(0,0,255) white: (255,255,255) black: (0,0,0) Source: wikipedia magenta(255,0,255)

13 RGB Color Image The dynamic range has some practical implications. 1. Say that your input B-band image has pixel values from 0 to a. If you remap such that 0->0 and >255, then you ll only see the bright stars (everything below 160 counts will -> 0). b. If you instead define the maximum to be some lower value (say 1000->255), then everything above this value will get mapped to and likely appear white if the object is bright in all three filters. 2. Astronomical images are generally (close to) linear. This may not be a desirable trait for making a color image. To increase the effective dynamical range (i.e. minimize the problem described above), it is common to apply a mathematical transform to the images before combining them into a color image. a. Common types of transforms to apply are a square root or a logarithm. i.e. sqrt(41000) = 202, sqrt(0)=0 log10(41000)=4.6...only works if all pixels >0 initially...

14 RGB Color Image Other things to keep in mind 1. The minimum pixel value (zero level). If you want to see the sky, this should be slightly below the mean sky level. If you don t want to see it, then it should be slightly above. 2. Gamma value: The gamma value is an alternate way or remapping the intensity scaling instead of taking square roots or logarithms. The basic transformation is so: response = (input intensity) gamma linear response: gamma=1 sqrt response gamma=0.5 In general, if you want to bring out faint features, use gamma<1. If you want to only see the bright objects, then use gamma>1.