MATTE PAINTING AND COMPOSITING

01/23/07

Assignment 01- paint texture on a sphere.

UV mapping
It is like sliding a picture on a paper. The picture sticks to the geometry. When u deform the geometry the picture will distorts. But when texturing, after the model is created the texture should fit right. And u have to know the difference between when u want distortion of the texture. For polygons go for a higher resolution geometry. For polygons you use the UV texture editor (correction tool), the UV’s are separate from the geometry. Here moving around UV’s can chg texture without deforming geometry. The UV space is independent of the geometry. You can prepare the UV layout before laying the texture or u can edit it after the texture is laid out. It has a flexible UV space. The textures in polygons are a projection. And you can also project a texture on the poly surface.

Nurbs UV space is married to the geometry. One cannot be changed without changing the other. It has a fixed UV space.

Projection mapping for Nurbs
To have the texture move with the object if the object is not deforming then just group the texture node with the object. But if the object is deforming then, one way is by creating a ref texture object. That is u freeze the texture into the UV space. (Texturing-create a ref object). This makes the texture into UV mapping. Another way is to convert the projection into a texture. This creates an image. In the hypergraph, edit-convert to file texture. It creates an image which is distorted according to the geometry so that the image when mapped it does not distorted. Maya creates a new shader with this image uv mapped to it. The same process works for polygons but the UV’s need to be laid out, if it has overlapping UV’s it can screw up.

Texturing a sphere
To avoid the pinch u can create an isoparm near the pinch so that it relaxes the UV’s.
The poles are depicted in a larger way. The map for a sphere would be wider than higher. That is the ratio of the diameter to the height is 2:1. We can create a texture for a spherical object. In Photoshop under filters use, under distort polar coordinates. Polar view is undistorted. Polar view is when u are looking at the sphere from the pole. It also a spherical view of a the polars flattened out. As in if the sphere is flattened or opened out from the middle. It creates a polar view for the top and bottom. So u create 2 of the polar views. If u took a sq and make a circle to fit it(1024x1024). Then u paint till the edges of the circle leaving some distance from the edge. Then u apply the polar coordinates. U get the top half of the sphere texture. Then you make the height half in the image size. Then in make the canvas size dble the height. That is the whole sphere. At the centre u can use the brush at the same size to paint. There should be no distortion at the centre.

01/30/07

To show the surface origins got to Display – Nurbs - surface – origins

02/06/07

Texturing the Rocking Horse
Use the light and shadow information in a photo to develop a good realistic texture. Multiply one image with the other to get more of a 3D look.
Histogram is a representation of the pixels in the image. The brightest point in ur image is white and the darkest point is black.

To make sure the material does not get distorted. Take the image size and make the projection that size eg. If image is 1536x1024 scale of projection would be 1.536x1.024
Planar project on the side first, and convert and close ,paint tex on that in photoshop. Then turn that projection to face the back or front side and don’t scale the projection, just place it properly and convert and close it.
Take the second image and put the whole thing on the first painted image. From the second image delete the stretched out part.
Then merge the top and bottom with the center piece either so that the seam is not there anymore or that it is more pronounced.

02/13/07

Perspective projection
This is what a camera does or what a projector does. Matte paintings use a lot of perspective projection. 2 and a half D technique used in matte paintings. A very simple shader with a little shift. Two and half technique requires camera projections. One of the important points is the camera settings. There are 4 attributes, to determine perspective, or the optical effect the camera creates.
Object size or image size from left to right.
Object distance
Focal length (camera)
Film back (camera

Focal length (camera) expressed through the angle of view. The angle of view is proportionate to the camera.

Film back (camera) the back of the camera where the light projects the image inverted. It is the physical size of the image. They limit the amount of exposure.

Nodal Point.
The point in the camera, where the light beams cross.


Film gate
Gives u the measurement of the film back.

Digital cameras have CCD’s for film backs. CCD stands for charge coupling device.

02/20/07

Nikon D200 – film back size is 23.6 x 15.8mm (0.9291”x0.6220” film back in maya)
Focal length is 200 mm
Size of the object or ball (Diameter) is 8.5 inch

Luminance shot allows u to see what kind of lighting is set up for the shoot. When u take the values or the range of the difference in the pixel colors then u can see how good or bad the lighting was.
Now use these values to compensate the lighting the picture. U use the actual image (inverted) of the luminance shot and place it on the ball images. Then u determine the difference between the brightest and the darkest value in the luminance shot. U add or subtract from the image. Use an over.

First compensate the brightness of the pics

In maya
First create a sphere
Create camera rename as proj camera 1
Display film gate
RG - Change image size to 600x400 (2:3 for still cameras) image as tiff
Save file

Change camera aperture to match the film back size i.e. 0.933 x 0.622 and film aspect ratio to 1.5
Change focal length to 200 mm
Bring in image plane of first image, change depth to 150, move the object left or right to match the image exactly. And move camera to fit the size of the ball in the image.
Lock the camera, put its pivot point in the center of the grid.
Move the ball back to the center. Then move the image plane so that it lies just a little behind the center of the grid.
Crop the ball images to exactly the size of the ball in photoshop. Rename it and bring it back into maya as the image plane.
Then adjust the film gate and adjust it to the actual ball so that the image will fit the actual ball. Bring the height of the film gate down to match the ball exactly and then make the width exactly the same.
Duplicate the camera, 9 times with 36 degrees rotation. Creates 9 cameras around the ball.
Create a surface shader and with projection create a file for its color. The proj change to perspective type and select the camera that matches the view.

03/06/07

Bump maps and Displacements
Displacements actually displace the geometry. It depends on the normals. Bump and displacement depend on only one channel value either 1 or 0. for a bump maya uses the luminance values to create the bump. You have to understand the macro and the micro details, also the range of the depth.
Large details will use a displacement and smaller details will use a bump map. Bump maps can carry a lot of small tiny details, whereas the displacement can also carry a lot of detail, it is very intensive to render.

To create the bump we will use a threshold mask. Also we will use clamping. We start by looking at the channels. The blue seems to have the most amt of details.
Duplicate the blue channel and name it mask. Use the levels to clamp the image that is make the image black and white with distinct borders. To do so we can use brightness and contrast, or levels but the best is to use the curves tool. It has the most amount of control. First check in the histogram where the black values are at and also the white values, then put those values into the curves tool. The purpose is to separate the black spaces from the white. Then using lasso erase out the seams. Then do the same for the white areas. Then select the black area and shrink the selection, and fill with black. Now u have a perfect mask with the detail edges. Then create 2 copies of the image one masking the black one masking the white. Then create another layer with just the outlines of the seams, and make that area darker to make the shadows look more prominent.

Then using that layer inverse it and create circular gradients in each cell for the displacement map.

Assignment
Make a bump map and displacement for the soccer ball.


03/20/07
Matte Paintings

03/26/07

Make surrounding in 3D and do a camera move so that the main structure come into the centre of the frame.

04/10/07

Colour key uses the RGB values to key whereas the blue screen or the green screen uses the channels. Channels are basically luminance information pertaining to a particular colour. RGB is additive, where colors add up to be white and CMYK is subtractive is just that u r subtracting light by adding colors.

To create a mask for an image of a tree which has very tiny branches, and tiny details. We cannot use masking coz we actually need an edge. So if we go to the blue channel, it gives us the best black and white (luminance image) for us to key. Duplicate the blue channel and work on this one to create a mask, a true alpha channel. Select a small area that u want to work on, so that u take into account the tiny details. Then u use the curve tool to clamp the black and white (like in the soccer ball). You must clamp such that u don’t loose out on the details. Make sure the blacks are not clipped too much. Once u have the desired result cancel it and apply it to the entire image. You want to make sure u want the edges to be a little soft and mainly a grey scale, specially the edges of the small branches.

So its an artificially created mask and we need to premultiply it to create a true alpha mask. Invert the image so that the tree is white now. Go back to the RGB layers, and then under select go to load selection and load the blue copy. This basically means that it loads the white area of that image i.e. the tree. Cntrl h hides and unhides the selection. Then u invert the selection and delete, this will delete the sky area and give us an edge of the tree to work with.

Reload the selection of the tree again and then go to border under select. With this we can select only the border pixels. So enter the number lets say 2 pixels, which will create a border selection of the edge and 2 pixels inside and outside of the selection. You then feather it by 1 pixel. Then u hide the selection again by pressing cntrl h. Now we will use the Hue/Saturation to change the color of these edges, make sure it is not too dark, and at the same time u loose all of the blue spill that u got from the sky. U can make the edges correspond to the color picture that is placed behind. So try and match it with the background.

Or a better way is to use the color balance tool to color correct those tiny areas which have a spill on them. It also gives you more individual control over the highlights, shadows and midtones, preserve luminosity is off. This will bump up the saturation. then using the hue/saturation tool u lessen the saturation depending upon the background. Then u will reload the mask selection again, select the border about 3 pixels and then feather it 3 pixels. Then go to the curve tool or the levels tools, and adjust the imgae so that there is a slight amount of highlight on it still.

Now we will blur the foreground or the tree so that the image looks a lot more real. The second reason is since the tree if it was really in the image would be defocused. Use Gaussian blur and make it about .8. Add some noise to the edges (color not monochromatic) and then give it a Gaussian blur again with lighter settings it will look more realistic.

Working with colored image which has high contrast points. Since no single channel is giving us all the information, we will have to use all the channels to create the alpha. Using the red channel we will first extract the highlights of the image. Using curves make the image darker and brighter so that u get the highlights.

Then use the green channel to further extract the Object. Under image go to calculations, identify the layers in the and do an add of the red and green channels

04/17/07

Additional Elements - Fog, Rain etc
If you record smoke or fog on video first shift it to 16bit
Video the biggest problem is compression. And it looses mainly color information, it preserves luminance information and there is no compression on the luminance. Becoz humans interpret luminance information better.

Z-Depth
The further things are the more milky or hazy they get. This is becoz of z-depth. It depicts picture information about the distance about points in the image. It is depicted by a grayscale value.

A spot light has a light intensity curve to modulate the fall of the fog in the scene.
Create a spotlight and make the cone angle about 170. Under light effects create a light intensity curve. Go into the intensity curve node in the attribute editor. With the light selected open the graph editor. The frames here stand for the sample distance. Leaving the first and last key frame delete the middle keys. Then u can manipulate the curve to obtain the desired effect. Now this curve depicts a distorted information. This is still a light. The light needs to be exactly with the camera.
The light intensity should be linked to a surface shader and then u assign the shader to the objects in ur scene. U will get a black and white image showing us the z-depth.

04/24/07

Color Correction
For the luminance you look at the curves and the levels of an image. When an image has less green it means the image is not warm.
To check a color cast use a grey surface coz the RGB is all the same. Use a grey sphere on a shoot to pick up the lighting in the scene, for a shoot. Color casts also effect the shadows, highlights etc, so one would have to measure dark and bright areas.
Use the color sampler to sample areas in the image (under color picker tool, photoshop)
U try to find the brightest and the darkest values in the image using this.
We will color correct with the curves tool first, and we will do this separately with each channel in the curve tool.

In the red channel, add points to the curve. Adjust the curve points, so that u get an average of the values u already have. Eg sample on is 192,174,192, so the first point we will make the output at 192 which is the avg.


What u r trying to achieve is that the colors as in RGB have the same value.
What we are trying to do is, get the RGB values the same at every level whether brighter or darker.