5.6.07

This is it - the last entry and update to my website. The presentation and defense are over, and went relatively well, but not perfectly. It's to be expected that the project wasn't at 100% completion and success, and it still isn't, but I didn't give up after those mile markers and major deadlines. Over the past week, I've used what little free time I could manage to continue working on this project to put together a better interface for an everyday user that I can be a bit more proud of. You can find this updated program on the Explore My Project page with the other smaller, more specialized programs. This new interface offers a menu of options for the user to choose (infinitely), and also allows the user to input a file name for an image to open. It is definitely something I had wanted to do earlier, but I just hadn't budgeted my time well enough to complete it before the presentation and/or defense. That was a fault of my own as well as a result of getting caught up in the issue of noise in my images for so long. It may be too little too late, but I am much more satisfied with myself and with this project being able to offer this program as part of the finished product.

I've also continued to look at different image formats and sizes, and I have yet to fully understand how the IplImage structure deals with the various formats and padding or alignment information. I have tested my program(s) on various formats including .png, .gif., .bmp, and .jpg. I have found success with various .jpg images including various sizes that are not all multiples of 4 (for the word size to avoid padding). However, when I have tried other image formats, or tried converting other image formats to a .jpg with an exterior program, I have not had the same luck. When the Bresenham algorithm resizes the image, the display becomes distorted. It is angled so I believe it is due to an alignment or padding issue, but I have not been able to account for it. I have tried testing the image width as well as playing with the widthStep (used in alignment), but have been unsuccessful. I sincerely hope that someone can look into this in the future because it is unfortunate that my program will not work on 100% of images tried and tested. If anyone has a suggestion or answer, I am open to ideas!

4.23.07

Hallelujah!! The night before I present, I have an epiphany! I no longer have noise showing up in my scaled image. It makes life wonderful and extensions possible! I have not figured it out with converting the data to ints and back to bytes, but the problem has been solved! (Finally, right?) Solution: split the image into three single-channel images (red, blue, and green) and scale each separately then remerge. The problem I was having with this solution before was the average function; when I typecast the data as unsigned chars (rather than unsigned ints), there are no longer any overflow problems. Chars are weird animals in C and C++. Referencing the data as unsigned chars left the bits of data alone, but the typecasting (indirectly from the function call) altered the data before averaging and overflow was occurring. But not anymore! It's a big relief, and now I'm actually quite excited to present tomorrow.

I promise to update the rest of this website as soon as the presentation itself is over. Check back soon for MANY updates on the Explore page and possibly for added pages.

4.22.07 (2nd entry)

I found something exciting with OpenCV tonight! I found a function where I could attach a "scroll bar" of sorts to one of my windows and have it automatically updating a value. I used this to adjust my scale that I am scaling the image by, as opposed to having the user press 1 or 2 to increase or decrease. It is MUCH more user friendly and looks great! While I still haven't been able to determine the source of my noise problems or conversion issues, I'm excited to have been able to add another nice feature to the interface!

4.22.07

Things should be winding down now, so close to the end of the semester, but it seems that I just can't get past the struggle of the extra noise in my images. I have tried a variety of techniques for converting the array of bytes (chars) to ints and back as well as for averaging, and I just cannot get over this hurdle. I'm scared since I have to present the project in less than two days (1:00 on 4.24!), but I guess I'll just have to do my best and explain my struggles.

If I split up the three channel image into three separate one channel images (blue, green, and red), I can get the image to scale pretty well, but I see the noise as a result. If I don't split and re-merge the image, I try to convert the data from an array of bytes as B,G,R,B,G,R,B,G,... to an array of ints as 0BGR,0BGR,0BGR,... where a single int contains all the information for one pixel (and four bits set to zero for padding). When I use this technique, my image turns blue. I have tried a variety of methods on the conversion as well as the averaging, and I haven't been able to determine why only the blue is showing up. It must be in the way the data is stored, but I thought I had that figured out... clearly I'm missing something here.

For better news, I have been able to take an image, scale it using Bresenham, and rescale it back to the original size using Bresenham. It seems that it would work pretty well except for the issue of the added noise in the image. We see double the noise when the algorithm is run twice on the image (once to scale, and another to return to the original size). I've also noted that returning to the original is a pretty decent image when the original pass enlarges the image. If the first pass of Bresenham scales the image down, especially with a very small scale such as .5, returning to the original image is much more difficult because pixel information from the original was lost or dropped when the image was scaled down. This seems pretty natural, but at least when returning from an enlarged image, Bresenham appears to do pretty well. Now I should figure out a mathematical way to determine how "nice" of a job it really does. I'm thinking about calculating the average difference between pixel values of the two images, but the noise that shows up will distort this calculation. It would sure be nice to figure out how to get rid of that. <sigh>

Well...time to get back to experimenting with conversion and averages again. Hope to see everyone Tuesday for the presentations!

4.15.07

Well, I have the Bresenham algorithm working to scale images, but I'm still having some trouble with noise showing up when I use the Bresenham algorithm to scale the image. I still think it is in the way I am averaging the pixel values, but I'm not sure how to "fix" the problem. I'm still working on that. I also have a feature added into the program that allows a user to adjust the scale by which the image is resized (currently by pressing 1 to shrink or 2 to enlarge). I hope to change this feature to use the arrow keys because it seems more logical and user-friendly.

I also want to add a way to save and open videos as opposed to just using the webcam. I believe that the OpenCV platform has some built-in functionality for this. I just need to spend some time searching and browsing the online manual for what will work for me and do what I want to do. If there isn't anything built-in, I will have a lot more work ahead of me so hopefully it's really there like I believe it is! I'm also still wondering about working with still images as opposed to streaming video. I struggled with it the few times I tried to play around with still images in the OpenCV platform, so I will have to keep looking for a better way to do that and what format is necessary to make that work well with my program.

Looking forward to finalizing some "fun" aspects of this project and moving forward toward graduation! Kind of scared, nervous, excited, anxious, and everything else all at once! What a busy time!

4.4.07

Feeling more relaxed today. After meeting with Dr. Pankratz, I made more progress on the image processing. I was able to get the image to resize with OpenCV function cvResize and display correctly. It can shrink or magnify the image which is an improvement. I was having trouble with this before with creating the header values (since it uses an IplImage format) and allocating enough memory to store the entire image (it worked before to shrink because it just didn't use all of the space). It was a big relief to see that working.

Now my concerns are turning toward how I allow the user to change the scale of the resized image on the fly. I want to create a tool (probably a type of scrollbar) for the user to adjust the scale as the video continues to stream. Another concern I have is why the Bresenham algorithm is giving "weird" data back and the image is distorted. I should look more into how the data is stored in the IplImage format. Making progress!

4.1.07

Happy April Fool's Day! I have good news, and it isn't an April Fool's joke! After working with John and Ted and playing around on my own, I was able to get something working in the OpenCV platform. It might be simple, but it can capture frames from the webcam and display them, as well as an altered copy of the image in another window. I found some a resizing function within the OpenCV platform and used that to test how the resizing would work and showed the resized image in the second window. I am now working on getting the Bresenham algorithm to work in that platform, and I think I'm almost there! There are just a few small changes before I should have something up and running to test the algorithm.

The next step would be to find out how to open a still image in that platform so I can test and compare quality of images (without trying to keep up with the speed of the camera's frame capture). With this, I want to find other alogrithms to use in the comparison, as well as ways to vary the existing Bresenham algorithm. I also need to create a method for the user to zoom in and out in real-time while watching the streaming video.

3.27.07

I had a great spring break. I received a job offer, and I'm excited to be moving forward with my future. I have decided to wait for Grad School for now; it isn't the right time for me. As for my project, I did take some time away from it because I felt burnt out, and I had been going in circles not getting anywhere. Dr. Pankratz and I met just before break to work on the DLL process. We were successful in creating a mini DLL that worked with a small sample VB program; however, I was not as lucky when attempting to translate this process to the actual project.

I also started working with John Moss and Ted Trisco (since we returned from break) on using the OpenCV platform for image processing in stead of the VB interface I had been working with. I have a LOT of work to do with this, but they have been very helpful and shown me a lot of information about where to look. I think this will take a lot of pressure off of me as far as coding is concerned because OpenCV is all done in C. My algorithms and comfort zone are in C++ so they should work together easily. I'm looking forward to getting the OpenCV platform to work for me, but I know it will take some time and work on my part. I need to set more time aside to work on this project. Motivation is an issue for me right now so I am going to try writing it into my schedule. Hopefully I can get back on track soon!

3.14.07

Well, I know it's been awhile. I have been struggling with a few things, both personal and project-related. As for the project, I've been working on organizing a tutorial or description on Bresenham and am making progress. My hope is to go through at least a portion of it during my walkthrough tomorrow. The other struggle has been creating a DLL for my C++ functions to be called from the VB program. I have tried a few different ways of creating them, asked for help from those more experienced with it, and am still stuck not quite getting it done. I will be meeting with Dr. Pankratz again tomorrow, so hopefully we can make more progress on this. This is a big road block for me since I find it difficult to test algorithms, compare variations, and see if they work in general with the picture data without using the real algorithms. Once this is cleared up, progress should be steadier. Look forward to more frequent (or at least more consistent) updates on my webpage.

2.25.07

Today I worked a bit more with VB 6.0 and my user interface. I made progress with learning how to use different controls. I still need some help with how to share data between the controls. I have asked a classmate for help, and if that turns out to be unproductive, I will ask Mr. Blahnik for more VB tutoring. I also wrote a short function in my C++ program for eventual use in converting the data from an array that I will get from the webcam to an array of "pixel" data that can be used with Riemersma's algorithms that I now have working in C++ 6.0, currently using arrays of ints. My next steps are to learn a little more about VB, make a list of questions or options to explore with the algorithms, learn how to save the scaled image to a file for later use, and start writing a tutorial on Bresenham (as a start for my presentation). There's a lot to do!

2.22.07

2.14.07

I am a bit frustrated with my progress this week. I am struggling to model Listing Three (the two-dimensional algorithm) in a program that will allow me to test it. I am going back to square one and starting with more of a general design. I hope this will help me understand the algorithm better. I plan to finish this before the end of the week. The next step after that is to try to use the algorithms to start adjusting data in the array of actual picture data in my other program that reads the picture and re-displays it. That is my goal for next week. Tomorrow I will be at WorkForce 2007 at Marquette University. I hope to meet some potential employers and make contacts in areas that I may be interested in. I will also be checking out the Marquette campus since I hope to be accepted to their Graduate School. It should be a fun day. I will get back to work on the project on Friday.

2.7.07

Dr. Pankratz, Dr. McVey, and I had another successful meeting this morning. After our discussion, we decided that I will add more output to my test programs to help guide the user (mostly Dr. Pankratz, Dr. McVey, and myself, but also anyone who chooses to play around with the algorithms in the future). These test programs will be made available on my website by the end of next week, and updated periodically throughout the remainder of the semester. After working with Mr. Blahnik and practicing with VB 6.0, with a little help from Ted, I was able to get a program to load a picture, read the pixel data into a one-dimensional array, and re-display it in another picture box. This program will also be made available with the test programs when it is further updated next week. I am feeling much more confident in my progress after today. I am hopeful for more success next week!

2.6.07

My meeting with Mr. Blahnik this morning was very productive. I am going to load a picture into a picture box and grab the pixels one at a time to put in an array because it is just a one-time thing; the timing shouldn't matter at this point because we are processing a still image. From there the array will "look like" the array I will eventually get from a webcam for a frame. This will allow me to write code that will process a still image that will transfer nicely to processing streaming video. Mr. Blahnik and I figured out how the colors are stored in the array and he showed me how to manipulate them and re-display the picture. I feel like I'm getting somewhere! Now I just have to put all the code together to see it in action and actually working for me.

1.31.07

I met with Dr. Pankratz and Dr. McVey this morning to go over a game plan for the next week. I will be looking at Listing One and Two a bit closer to try to understand how they relate to Bresenham by graphing the results. Next week we will meet again to discuss the two-dimensional algorithm of Listing Three. I look forward to this because I haven’t been able to follow it so far. I will continue to study it and try examples, but hopefully we can get farther when we work together on it next week.

1.30.07

I met with Dr. Pankratz today to discuss the one-dimensional algorithms proposed by Thiadmer Riemersma in his article in Dr. Dobb's Journal. I feel I have a better understanding of how they work. My plan is to try to write some mini-programs to test the algorithms so I can see them in action to get an even better understanding of the process. I am also going to look at the two-dimensional algorithm and start looking for how to load an image from a file so I can get started on that portion of the project.

1.25.07

Mr. Blahnik explained how to use the Logitech 4000/5000 webcams today. He showed us his software and how to use it to capture frames and streaming video. It was useful to learn how the raw data comes to us from the camera. My plan is to study his code to learn a little VB as well as how to manipulate a still frame (to which I can later add a loop to handle streaming video). I also worked on my website design and my personal philosophy and career path statement.

1.23.07

Today I met with Dr. Pankratz to receive my project description. I am excited and nervous about the project: image scaling in real-time. My first step is to read the articles about the smooth Bresenham nearest neighbor interpolation method in Dr. Dobb’s Journal and understand the algorithm. I will meet with Dr. Pankratz and Dr. McVey next week to go over the algorithm and an example to be sure we all understand the process.