Elizabeth McMahill

Freelance Artist/Animator



Experiments (Lessons 1&2 current)

Before I get to posting, here's a tip I'm not sure if other's have realized - if you're on a machine running windows and can't open up your glitched images in photoshop or the gimp, but don't want to rely on taking screenshots, try opening up your edited images in Paint and resaving it out from there.  It's quick and looks to match the interpetation of picture viewer.

PART I - Lessons 1&2 - Hex editing, Gifs, and opening things the wrong way


To start things off I grabbed my favorite image from the internet and started playing with it in the hex editor (I'm on windows currently).  I've did my fair share of hex editing of images a couple of years back when I got into this area - but the idea of search & replace to edit images is a new one, so I mostly played with that.  One particular thing I found interesting was that you could search and replace without being case sensitive - for example by replacing "a" with "a" "A" would be replaced as well - essentially characters could be clamped to uppercase or lowercase.  This sort of change only subtly changes meaning in the language for which it was intended but of course would drastically change the images they were being used to abstractly represent.

Here, only the letter P was "flattened".

Here, the letters B, P, E, and A were all "flattened"

Search and replace was tried on this image of corn as well - resulting in those distinct horizontal stripes I have seen before in other images but had never before encountered myself.

In this last case I "flattened" the character "c" throughout the image.  The striking interplay of patterns from both the original and corrupted images is something I haven't seen a lot of but would very much like to explore more in the future.

From there I went on to investigate what visual changes would occur when a png file was altered.  The results were highly interesting but also highly, and invariably, noisy.

Both interelaced

And non-interlaced.


Reading through some of the projects that had been posted last Friday morning someone mentioned using photoshop raw for manipulating images.  I had never used the format before in any context and decided to give it a whirl.  I didn't use it to corrupt/glitch any images proper but instead was fascinated by the options given for interpreting (or misinterpreting) images.  Very interesting effects resulted from supplying incorrect dimentions, depth, and channel information.  It's also a quick way to visualize raw data from anything else - all you have to do it open it!  Since my pinapple image started with high compression and a rather small file size I opened it up in photoshop, saved it out as a photoshop raw, and then tried re-opening it under various parameters.

The first few images with default parameters.  In an attempt to get something that resembled the orginal I set the dimensions to match the starting jpg.  However, I left it on the 2-channel setting which resulted in pretty interesting interpretation.

8-bit values

16-bit values.

From here told photoshop to look for all three channels, but still let it guess at the dimensions.  It gave me these wonderful offset images.

Interesting too was asking photoshop to interepret the image as not interlaced, when the file was saved with interlacing, or vice-versa.  The last image shows the effects.

After having messed around with images I wanted to see what non image data would look like when opened up as a photoshop raw.  Since I needed a reasonably large file I went with the actual photoshop.exe file, copied it over into my working directory, changed the extention, and went on to opening it up inside itself.

This is what Photoshop thought it looked liked.  2-channel grayscale image with guessed size.

3-channel rgb image with guessed size.

3-channel rgb image with tall/skinny dimensions, 100px wide.

A close-up of the strip

And a gif that scrolls down the strip.  Frames were exported using photoshop actions to save on time, but percentages were used when cropping down the canvas size, so scrolling misses data at the beginning and then slows down at the end.


PART X - Earlier experiments and observations

In addition the images made in this workshop I wanted to share some earlier and recent experiements and observations that I've made.  Maybe these will help broaden the range of this course experience and help spark ideas amongst ya'll - something that this course has been exceptional at so far!  It's been super cool to see everybody's work and all the different ways thought up to approach images!

Video game glitches have already been shared and are pretty popular.  I thought it was interesting to pair it with realworld patterns like the wave & moire shadow pattern from a window screen on vertical blinds since they both exhibit abrubt blockiness and subtle patterns and are "naturally" occuring.

I never had cable at home so the swtich from analog to digital signal was very noticed by me and still proves fascinating after a few years.  It makes bad signals so much harder to watch but at the same time is so much more visually interesting and sudden.  It's a great "natural" glitch.

It's also pretty close up.

The following images aren't glitches per se.  However, conceptually, one of the most interesting aspects of glitch art is that new meaning is created through re-contextualizing and altering existing data.  I see hex-editing, datamoshing, etc as most fundamental ways to do this, but not the only.  The other part that draws me to glitches is that through the destruction of image you come to better understand the true structure of image.  The following deal with these ideas while departing from traditional glitches - though they still produce similarly interesting imagery.

These two images were created using the wave filter in Photoshop (Filters>Distort>Wave)

The following is a :30 video displaying interaction effects/artifacts/errors/glitches in after effects caused by spinning three squares (r, g, and b) in on themselves at high speeds with motion blur.  The inherent quantization of digital processes and video is what leads to calculation "errors"/innacuracies and creates interesting edge behavior and patterns as a result.


This one was just a black image with a white square in the lower right corner.  It was saved out as a jpg, re-opened, enlarged, and adjusted to bring out the color and texture of the induced compression artifacts inherent in the format.

This picture was to explore jpg compression too, but from a mathematical stand point.  I was listening to the Coursera Image and Video Processing lectures (https://www.coursera.org/course/images the class is over now, but hopefully they'll offer it again in the future, very eye opening) and learned how the DCT was used in jpg compression (http://en.wikipedia.org/wiki/Discrete_cosine_transform).  The DCT per image block is what results in all those interesting tiny patterns.  To get this image I saved out a found image with very high compression and then re-opened and heavily edited/overlayed the result to bring out the patterns specifically created in the compression process.  In the top middle there's a little sad face!

This last one is similar to what I did in Part I by opening the phoshop executable as a photoshop raw, but instead it takes the data from notepad.exe and displays it as a bmp image.  There was no middle software involved in this, just a calculator, hex editor, and a bit of research.  I wrote a little more about it on my blog, http://elizabethmcmahill.blogspot.com/2010/05/notepadbmp.html  Making an image like this isn't very easy, but it's pretty cool.  There's some tedious calculations and conversions that need to be performed to make the file work.  It'd be best if they were executed by some program, but they're still very do-able by hand.  If you want to learn about the bmp format and how to write a header for it check out, http://en.wikipedia.org/wiki/BMP_file_format It gets kind of dense, but there's some examples further down that help make sense of it all.


Please sign in or sign up to comment.