Goals
Learn a little about photography Understand how basic camera systems work
Recognize intelligent features Show how neural nets and fuzzy logic are used in digital cameras
Outline
Brief history of the camera How a digital camera works Challenges in digital photography Intelligent features in todays digital cameras Case studies
Began to rival wet plates Pros: photographers could use commercially made dry plates instead of preparing their own emulsions in a mobile dark room Allowed cameras to be small enough to be hand-held Shortened exposure times lead to invention of the shutter
1925 Leica 1
Introduced 35mm film
Opens to expose the film or sensor to light Closes to stop exposure Shutter times measured in seconds
i.e. 1/100, 1/50, 1/20, 1/10, 1/5, , 1, 2, 4, 8 Also measured in stops of lights
Sensors
CCD charge coupled device, analog shift register CMOS - complementary metaloxidesemiconductor digital recorder Requires less power, less susceptible to noise
combination of shutter speed, aperture, and film speed (ISO) a combination of shutter speed, aperture, and ISO that provides a correctly exposed image Underexposure too dark Overexposure washed out
Correct Exposure
Exposure
For any given situation, there are many creatively correct exposures Seven different kinds of creative exposure
Story telling small apertures (f/32, f/22, f/16) Isolation large apertures (f/2.8, f/4, f/5.6) Who cares in between apertures, depth of field isnt a concern Macro out of focus backgrounds (bokeh) Freeze action fast shutter speeds (1/1000, 1/500, 1/250) Panning medium shutter speeds (1/60, 1/30) Imply motion snow shutter speeds (1/4, , 1)
Exposure
A Stop of light a doubling or halving of light
Can trade one stop for another to keep the same exposure Sunny f/16 rule: on a sunny day, f/16 and 1/ISO to give correct exposure
File types Auto Focus Auto Exposure Noise Reduction Sharpness Contrast White Balance
Pros used everywhere, small file sizes, good compression, fit more on a memory card, faster download times Cons lossy compression leads to data loss and image degradation from resaving Pros lossless image compression, no image degradation Cons not universally supported, large image sizes, still lose some information and ability to process photos
Pros full control over image processing, smaller files than TIFF, completely lossless, no compression Cons almost no support, formats constantly changing, must do all processing manually
Active
1. 2. 3. 4. Sends a signal towards target Measures returning signal Calculates distance Drives lens to desired position
Passive
1. 2. 3. Looks at current image Move lens slightly, compare the two images (sharpness, contrast) If better, keep moving. If worse, go other direction
Metering Techniques
Spot only measures in a 2-3 degree cone in the center of image Center Weighted weights the center 50-60% of the image Average averages the entire scene Partial only uses a certain region (usually the center) Evaluative uses information from several zones and combines the info to find the best exposure Scene Modes gives the camera an idea of what it is seeing so it can apply special rules when metering (night, indoors, museum, snow, sports)
Intelligent Cameras
Todays digital cameras have many intelligent capabilities
Neural systems
Evaluative or matrix metering Face detection
Intelligent Cameras
Auto Exposure Decision Making
Once an exposure has been calculated (using any of the metering methods) must then choose the creatively correct exposure Fuzzy rules based on
Available ISOs, shutter speeds, and apertures Weights or priorities given to different settings
Intelligent Cameras
Focus Point Selection
Most newer cameras have multiple focus points Must choose which focus point you intended the camera to use Fuzzy logic rules:
If 1 is near and 2 and 3 is far, 1 is somewhat likely If 1 is near and 2 is near and 3 is far, 1 and 2 are highly likely If 1 and 3 are far and 2 is near, 2 is highly likely Ect.
Image is compared as lens focuses to choose the best image Using a sharpness measure
Define over a portion of the image Use fuzzy membership functions and rules to drive the lens
Fuzzy rules
If sharpness is increasing fast, move lens fast in same direction If sharpness is increasing slowly, move lens slowly in same direction If sharpness is not changing, do not move lens If sharpness is decreasing, move lens lowly in opposite direction
Case Study
Face Detection via neural network