LumoLabs: Busting the alignment myth

This will be a special feature for hard core falconeye followers ;)

It all started with the exciting new feature of the Pentax K-7 camera which can produce true HDR images in-camera (it takes 3 shots, 3 exposure values apart from each other, and creates a single, tone-mapped or blended image from the sequence). It does it so well, esp. in the medium setting (as opposed to strong) that a limitation of the feature does actually hurt: you require a tripod to use the function.

The general opinion why Pentax left out the alignment step from their HDR function is that it is a processing power-hungry computation not easily done in-camera.

Not my opinion, though. I felt that Pentax has left out 10% of the implementation which reduces 90% of the functionality. Shouldn't it be the other way round?

Now, opinion isn't proof.

So, I implemented my own alignment operator, benchmarked it and compared its quality against the state-of-the-art in HDR.


The algorithm

The implementation is based on an algorithm whose computing time increases linearly with image size and is constant with amount of distortion. Note that I did not attempt to cope with the alignment problem known in panorama stitching, i.e., I don't do SIFT key extraction, I don't correct for lens distortion, I don't project onto a sphere before alignment, etc. pp. However, what I do is correct for shift and rotations which is a mathematical first order approximation to the latter and rather exact for small shifts and rotations.

The implementation uses a monochrome quad tree and high bit depth. The coding is in Java. It isn't based on anybody else's intellectual property.


The benchmark

I benchmarked the algorithm with a 5 image HDR from Munich Siegestor. It had distortions of up to 20 pixels and needed rotation as well. The source images haven't been ideally sharp (e.g., 1/25s and wide open with 15mm). And no program created a perfect alignment. So, I considered this example be adequate. The images were 14.6 MPixel JPG images.

The processing times on a Mac Mini (Early 2009), one processor used, are as follows:

• Single alignment: 175 ms
• Alignment of all HDR images: 700 ms
• Reading of all images from disk into Java heap: 16 s
• Shift-turn and write images to disk: 34 s
• Total, disk to disk (70 MB transferred), 5 images: 50 s

The K-7 has a dedicated image DSP and can certainly rival this. E.g., it can JPG-compress a 14.6 MPixel image in about 100 ms only!

Because the HDR function within K-7 already has created the required image structures in its 2 GB memory, it would be conclusive to assume that addition of an auto-alignment feature to the HDR function would add less than an extra 1/2 second to overall operation time.

If processing speed isn't the reason, so maybe it is quality? This must be looked at. In particular as some of the alignment operators I compared with took several to nearly 10 minutes to complete.

The quality

I've run several alignment operators and come up with the following ranking:

1. Photoshop CS3; PhotoAcute ("normal alignment")
2. Falk Lumo Operator (as of this blog article); PhotoMatix 3 ("by matching features")
3. PhotoMatix 3 ("by matching features"; "reduce ghosting")
4. PhotoMatix 3 ("by correcting horizontal and vertical shifts")
5. No alignment

All pre-aligned images have been read into Adobe Photoshop CS3, either by HDR merge (without further alignment) or by opening the 32 Bit .hdr-file written by PhotoMatix. Tonemapping was done in Photoshop CS3, using the "Equalize Histogram" method. The latter isn't best but as it has no parameters, I thought it would make for a good option in a comparative study. Finally, the result was studied to derive the above ranking. Which is entirely subjective!

Below are the two samples from category #2. Please, visit the gallery to view the other samples, too. Note that all sample images have been resized to 6 MPixel.

Fig.1 No alignment


Fig.2 PhotoMatix


Fig.3 Falk Lumo Alignment Operator (as decribed here)

Take me to the gallery

Conclusion

The myth is busted. There is no reason to not include auto-alignment into the next firmware release of the K-7 HDR feature. In less than an extra second of processing time, about PhotoMatix quality can be provided. Make it so!

K-7 Barbecue

K-7 Grillen

In order to celebrate the end of the Pentax K-7 alpha test, another tester from Munich and myself organized a little barbecue.

I've cut a little movie from files found on my K-7 and some shots from friends (in German language).

K7Grillen (K-7 Barbecue) from falconeye on Vimeo.

Enjoy :)


Btw, the green light at the end is the K-7's AF assist light.

Pentax K-7 Concluding Review Report

Fig. K-7 getting wet feet while recording a movie (aperture blades stopped down).


This is my concluding review report about the Pentax K-7 dSLR camera hitting stores next week (last week in Japan).

It is based on my experience throughout the past month, after Pentax Germany was so kind to ask me to test their new flagship camera. It is today that I have to return the camera :(

You may read about my detailed findings in my various blog articles found here:

• I am Pentax K-7 Alpha Tester :)
• In the rain
• Its continous autofocus -- Part I, Part II
• Its precision autofocus
• Tungsten high iso noise
• Comparative noise study K-7 vs. K20D
• Comparative resolution study K-7 vs. K20D
• As a movie camera -- Part I, Part II, Part III
• Ergonomics

Again, I am not going to repeat the spec sheet of the camera. Please, go to

• pentaximaging.com K-7 website
• K-7 user manual
• DPReview web preview/review

to get the full list of features. Here, I'll just share my opinion about the product. First, I am going to break up my impression with various "use cases" in mind.

1. General comments

The Pentax K-7 is in a class with only five competitors (*). It is Nikon D5000, D90, D300, Canon EOS 50D, and Pentax K-7. The price range covers 730$ (D5000), 950$ (D90), 1200$ (50D), 1300$ (K-7), 1700$ (D300). However, as the K-7 is at launch price now, it is expected that the street price of the K-7 is to drop below 1000$ later this year and would be priced roughly like the D90.

So, let me quickly compare the K-7 against the other four models:

• K-7 vs. D5000
  Pro: Better LCD, Faster/better shutter, much better VF, SR, WR, metal body, 15MP
  Con: Heavier, no tilt LCD


• K-7 vs. D90
  Pro: Faster/better shutter, slightly better VF, SR, WR, full metal body, 15MP
  Con: none


• K-7 vs. 50D
  Pro: Slightly better VF, video, SR, WR
  Con: 1/180X, no double cross center AF sensor, no tethering


• K-7 vs. D300
  Pro: Lighter, video, SR, 15MP
  Con: slower and no 150k shutter, 1/180X, no 51 AF points (11), no 1005 zone metering (77), no tethering

_____
(*) defined by dSLR, 12+ MPixel APS-C, LiveView, ISO3200+, 3.5+ fps
SR: in-body stabilization (shake reduction)
WR: sealed body and kit lenses for weather resistance

This little list is not meant to rank the competition ;) But it shows that the K-7 is very interestingly positioned. Let's say that the D5000 seems to be an interesting camera but is positioned below the D90 which is a more serious competitor, but still somewhat below the class of the K-7. The only true matches in class are 50D and D300, where D300 actually is somewhat above. However, if video or in-body stabilization or bad weather-resistance are required then there simply is no match for the K-7. The only thing we need to check out is whether K-7 delivers on its promise. If it does then it has no real competition (yet).

Having said this, I'll leave the feature list comparison for a more meaningful consideration of strong or weak fields of application. Each of the following fields gets a score within 1 to 5 stars (category for digital 35mm mount SLR).

Throughout, this is based on the following judgements:
- High ISO noise performance is on par with the current state-of-the-art for APS-C (ignoring differences of a 1/3 stop). The K-7 will look both more noisy and more detailed at the pixel level compared to the competition above. Some think that this renders more beautiful images when printed. Others think that buttery images at high ISO are better. I think that the differences aren't significant enough to be important but I prefer to keep noise and detail for later processing. Full frame delivers one full additional stop though.
- Autofocus is significantly improved compared to previous models from Pentax and at or above average. But still no match for the best in its class (D300).
- 15 MP resolution is fully applicable when using good glass.
- Everything what you can read in the separate articles as given above.


2. Landscape / Cityscape photography
★★★★★
Actually, the K-7 is a stellar performer here and K20D already scores high. The relatively lightweight body, stellar image quality at low ISO in combination with available stunning prime lenses makes this the best choice short of a (heavier) full frame 20+ MP camera. A function for automatic horizon just makes it perfect.

3. Wildlife
★★★☆☆
It scores high like for landscape, but I remove two stars for the following reasons:
- fast and/or long glass is extremely rare and expensive, and a tele converter is lacking.
- AF.C autofocus is good (and much better than in previous Pentax cameras) but may still miss a shot in action-loaden situations (flying birds, predator attack).
Note that the first point is not a weakness of the camera itself.

4. Sports and Action
★★★☆☆
Same rationale as for wildlife. Also, there are faster, machine gun type bodies.

5. Wedding
★★★★☆
The K-7 is the perfect APS-C body for this kind of shootings and additionally can provide stunning HD footage. It is a robust tool and sufficiently fast, too. However, in situations where light is not optimal, a full frame body with fast glass can be a better option.

6. Family
★★★★☆
Much like wedding. But ease of use and the fact that full frame would be overkill adds a star. Very young active kids or pets can still be challenging though and costs a star.

Btw, I don't agree with some other reviewers that the K-7 underexposes. If I would, I would subtract another star here. I agree that K-7 tends to expose darker than what is seen in other cameras, but for a reason. Easily studied with its live histogram. E.g., a burned highlight in the extreme quarter decreases exposure from 1/25s to 1/30s. So, it isn't true that K-7 underexposes to rescue irrelevant highlights. It just takes into account that bright levels clip where dark levels don't. Just work with +1/3EV if you never adjust levels.

7. Street and People Photography
★★★★★
Can't be beaten. Light-weight un-obstrusive and able to cope with almost every foreseeable situation. A smaller camera (like K-m or non SLR) may be interesting but would be less rugged. The improved metering performance with highlights and with flash are important improvements coming with the K-7 (for night life).

8. Travel
★★★★★
Same rationale as above. Just keep an eye on your stuff, though. An SLR may not always be the best option. Depending on your priorities.

9. Portrait and Studio
★★★★☆
The K-7 is almost perfect. The lack of a tethering option costs a star, though.

10. Movies
★★★☆☆
I judge the K-7 to be the second most interesting SLR offering video (with Canon 5DmkII being no.1 which would get 4 stars). The missing 2 stars are:
- No supersampling of full 15MP down to 1080p
- No videographer form factor/controls/EVF available with working Contrast AF
(taking the strengths (ISO, DoF etc.) into account.)

What sets it apart from other offerings are a "better than 720p" mode @30fps, uncropped 3:2 video, stereo, bearable jello effect, some manual control (aperture, EV compensation, exposure lock), no motion-compression artifacts. Of course, full manual control, electronic viewfinder and a fast contrast AF are all missing.

11. HDR and panoramas
★★★★★
The built-in HDR feature makes any other camera score 4 stars or less only. But Pentax missed the chance to align the images in-camera which would have been straightforward to do. The feature works great, otherwise. Moreover, the K-7 offers perfect control to do HDR panoramas. Live view with grid, exposure bracket and horizon control are an additional help.

12. Macro photography
★★★★☆
Perfect control available. A higher resolution LV would be an improvement but doesn't currently exist on the market. I subtract a star because an automatic focus series for focus stacking would be great. AFAIK, it doesn't currently exist on the market for no reason.

13. Astro photography
★★★☆☆
LiveView is of great help and controls are perfect. Missing though:
- Tethering option.
- Dark frame subtraction shut off for exposures > 30s not possible.
- Crop of FoV compared to full frame, depending on the type (FoV) of telescope.

14. Final Verdict
Four stars is the average and my conclusive score. If you look at my criteria, there probably will be no camera scoring at 5 stars (smarter beats bigger). Certainly not its direct competitors, the 50D and D300. I like both cameras as well and they have different weaknesses and strenghts. But I like the K-7 better. A contender for 5 stars would be Nikon's D3X, if it were 1/3 lighter and had video.

So, the K-7 not only feels like a lot of fun to use, it is a very good proposition on the current market. The best ever made by Pentax and a bigger step forward for Pentax than they did whith the (already very good) K20D. Compared to the K-7, the K20D feels old already.

So it does deliver on its promise indeed which means:

Pentax K-7

★★★★☆

Editor's Choice

"APS-C SLR camera"

(5 stars not assigned in category; as of 2009, July 1st)