A study commissioned by the South China Morning Post estimates 140,000 people took part in the July 1 march.
The result - calculated through computerised area-density analysis - is similar to separate estimates by two University of Hong Kong academics but falls way below the organisers' figure of 510,000. The police did not offer a total but estimated the largest number of marchers at any one time at 98,600.
Watch: How we got our numbers for the July 1 march
The number of people taking part in the annual rally has always been controversial, with police, the organisers and academics releasing widely different figures.
The dispute this year is further complicated by the slower-than-usual pace of the procession. In 2004, it took a protester on average of 90 minutes to march from Victoria Park to Central. On Tuesday it took about three hours.
The Post commissioned chartered land surveyor Thomas Lee Wai-pang to provide an independent and alternative analysis. Lee and his team used "the density and area computation" method that has been used by other media organisations to calculate crowd size.
Next he calculated the average density of the crowd by analysing photos taken by Post photographers at three different locations throughout the event. The photographers took pictures of the crowd at a near vertical angle from 30 metres above ground at locations in Causeway Bay, Wan Chai and Admiralty.
By analysing the photos with the satellite image analysis software, Lee was able to establish the average density of the crowd at the three locations at different times. Multiplying the average density with the total area of the route, he then calculated how many protesters were filling the entire 4km route when the people at the front reached the end.
To do this, three Post reporters were assigned to the task. The first left Victoria Park at the head of the crowd at 3.25pm. He reached the finish point in Chater Road in Central at 6.20pm. The second reporter then left Victoria Park and arrived in Chater Road at 10pm. By then, the third reporter, who was at the tail of the procession, was in Wan Chai. This means between 6.20pm and 10pm, only half of the route was completely occupied.
Based on the density, Lee calculated that 91,213 people had taken part in the rally between 3.30pm and 6.30pm, and 49,195 between 6.30pm and 10pm. The two batches add up to 140,408.
The result was surprisingly lower than the public perception. Most people believed Tuesday's crowd was comparable to that of 2003 or 2004, when half a million people took to street.
This could be partly due to the extremely slow pace of the rally and the crowd's unusually high density. The average density in Lee's past studies was 1 person per square metre, on Tuesday it was 1.4.
Police earlier accused the organisers of deliberately moving slowly, while they said the police caused a bottleneck by not opening more lanes for marchers.
"It was unfortunate the crowd moved too slowly. The slow pace might have discouraged some to stay on and finish the march, which could have lowered the turnout," said Professor Paul Yip Siu-fai, a social statistics expert at the HKU.
The Post acknowledges some limitations in the study. Some marchers who joined or left the procession in the middle of the course may have been excluded for instance.
The stalls set up by different groups along Wan Chai and Causeway Bay also mean the actual route area could be smaller, suggesting the number of people could be even lower.
Yip said he found the Post's study was "sound" as it tried to capture the varying crowd sizes at different locations.
The Public Opinion Programme of the Hong Kong University refused to comment on the analysis.
Edward Tai Chit-fai, senior data analyst of the programme, said computer counting should be separately verified by manual counting.
Programme director Dr Robert Chung Ting-yiu earlier put their estimate at between 154,000 and 172,000. Chung said his figures might not align with the public perception because people had been affected by "imprecise figures". He called on the organisers to conduct the estimate "scientifically".
The organiser, Civil Human Rights Front, welcomed other organisations to do their own counting.
"We believe our count is already a conservative estimate," said front convenor Johnson Yeung Ching-yin. "Some people might have left in the middle of the march and it is almost impossible to determine the number of these people."
The police yesterday refused to comment.
Additional reporting by Tony Cheung, Fanny Fung and Emily Tsang