Sandbox‎ > ‎Archive‎ > ‎IPT 2008-09‎ > ‎Blair's page‎ > ‎Major Project Register‎ > ‎

2009-07-08 Information processes (image) + miscellany

posted Jul 7, 2009, 10:49 PM by Unknown user   [ updated Jul 16, 2009, 11:39 PM by Eddie Woo ]

For the record

For the record, Mr. Woo confirmed yesterday that print advertising is suitable for the text component of the multimedia project.

Furthermore, during morning class, Mr. Woo said that my video could not be uploaded to any website other than Google (the intranet runs on Google Sites), as my release form specifically stated that I would only upload to the JRAHS intranet. This is true even I upload the video to a website and make it accessible by JRAHS intranet users only. I therefore cannot rely on third party services to host my video (e.g. YouTube, PWA, or VoiceThread).

How I created panoramic C-block photos

I have spent the past week "stitching" many image files together to create panoramic photographs, as I had described in 2009-07-01 Updates. The purpose of this process is to produce a single image that has an unusually large angle of view.

Shooting for panoramic stitching

The source images were collected by a digital still camera which organised them as JPEG files and stored them to a memory card for later retrieval (as described in 2009-07-01 Updates); this is a screenshot of the files being displayed as thumbnails in a file manager on Windows 7:

The observant will notice that the photographs were taken in portrait mode instead of horizontal:
  1. One normally aligns the longer side of the camera with the dimension that needs to be maximised.
  2. This dimension is usually width for landscape photographs and height for portraits, hence the term landscape mode for images produced by maximising the width, and portrait mode for images produced by maximising the height (a naming convention that has spread to print production in general).
  3. In this instance, the width of the final product could be increased simply by taking more photos while shooting (remembering that these photographs were taken for the purpose of panoramic stitching), so it made more sense to try to maximise the height (which could not have been increased later)
This technique is aided by a built-in orientation sensor in the camera that identifies the positioning of the camera in relation to the ground. The data collected by this sensor is analysed and added as metadata (specifically, EXIF data) to photographs. As a result, I never had to manually rotate the images shown in the above screenshot - the system did this automatically using the embedded EXIF data.

Putting the pieces together

The images were pieced together in Adobe Photoshop, which analysed the individual images by searching for objects (window frames, doors, handlebars, etc.) that were common and then sorting these images into separate, labelled layers. These layers were then cropped and manipulated so that the final image would be smooth and continuous (albeit a very distorted representation). Photoshop refers to these steps collectively as "Blend Selected Layers Based On Content":

Cleaning the edges and saving the result

Due to the amount of distortion that results from this highly complex process, the edges of the image produced are not straight, but instead very rough. It was therefore necessary to crop this image (yet another example of processing):

Even after cropping, though, the image produced was 17182 by 3281 pixels in resolution. This is far too big for viewing and would take too long to display, so I resized (yet another instance of processing) the image to 6284 by 1200 pixels, deciding that 1200 pixels would be a good "standard" height for panoramic images. I then saved this image as a JPEG, an example of organising:

Processing to compensate for exposure

Photographs for stitching must be taken under very similar exposure conditions, or else there will be very obvious colour changes throughout the image, which looks very bad and unprofessional. Exposure conditions can be kept similar by using constant aperture, shutter speed, "ISO" sensor sensitivity, and white balance.

However, this results in problems when there are two areas with very different lighting. For example, one area may be lit by dim artificial lighting, and another area by bright natural sunlight. If these two areas are taken with the same exposure settings, one area will be underexposed and/or another area will be overexposed.

This affected one or two of my photographs, so I opened them in Adobe Lightroom to add a virtual graduated filter. This changes the exposure of the image gradually, and is therefore perfect to my "partial overexposure" problem:

Professional photographers with the right equipment have used real graduated filters for many years, and using a real graduated filter would be part of the collecting process; however, because I have simulated the use of a graduated filter using software, it is actually an example of processing.

What next?

I now have 11 images produced in exactly the same way as the worked example above; all of them are 1200 pixels in height and a few of them have been fixed in Lightroom. I will soon be distributing them through the Intranet, an example of transmitting and receiving.